I have a question regarding the new third gen torque sensor from Sempu. Specifically does it require a power stroke that is perpendicular to the chain line like the tdcm or does it function more like the thun sensor where the power stroke can be from any direction. This relates to being able to use it on a recumbent.
Sempu t2.3 torque sensor
- Thread starter Hangdog
- Start date
Hangdog said:
Hey Hangdog, the Sempu sensors aren't like THUN or TDCM in this regards. They more or less put out a signal whenever there is any kind of force that bends or stresses the spindle. It doesn't matter if it's a forwards torque, a reverse torque, or you just standing on the pedal at the bottom of the stroke putting no torque or power into the drivechain at all. That still produces a voltage on the output.
In practice though, when you are riding an ebike the amount of general stresses on the spindle increase in proportion to how hard you are pedaling, and so it still results in an effective behavior for proportional assistance.
The THUN/NCTE sensors actually measure torque proper across the spindle. If you twist the spindle it puts out a signal, while if you put sideways or bending forces on the spindle it doesn't register.That allows them to isolate human input power in watts (= torque * rpm) with decent accuracy.
The TDCM sensor measures how much the right side of the spindle is pulled backwards as a result of chain tension. That can be calibrated into a decent human power readout if the front chainring diameter is fixed, but with a 3 speed front chainring it can only be calibrated for 1 of the 3 gears. If the human pedal force isn't perpendicular to the drivechain like on a recumbent it will still work OK. The gain from the right pedal will be less than that of the left pedal stroke because the pedal force forwards partly cancels the chain pull backwards, but you can just calibrate it accordingly and it will still work fine.
thanks justin,appreciate your detailed reply.So I infer from your explanation the sempu will work on a recumbent setup then. As far as the TDCM sensor goes, my personal experience with installing it on my recumbent trike was it didn't function at all. no matter how much I adjusted it, it didn't send any signal back to the CA3.After exchanging it for a thun,it immediately worked perfectly. As the sempu unit seems very well priced I thought It might be a good choice on my next project...thanks
Hangdog said:
Correct. I could have been a bit more explicit about that
Yeah it seems to hit the right mark on almost all aspects we'd want for pedal assist control. Low price, very high pole count for almost immediate response, both left/right pedal force sensing, and a decent signal range (1.5V to 4.0V). The main downside is that it also requires drilling a fairly large hole in the bottom bracket of the bike to fit the HiGo plug.
The earlier generation of Sempu sensors had the cable coming out the side of the device which was awesome as no modification to the bike frame was required, but those had other reliability issues. We're hoping these units are more rugged and first impressions seem to suggest as much.
madin88 said:
Note that Sempu offers two options for the PAS encoding. Either 48 pulses on one signal and a FWD/REV direction indicator on the other signal, or quadrature style (sin / cos) encoding with 24 pulses per rotation on each signal. The latter "double pathway" option as Sempu calls it plays nicest with CA3 firmware as that's how most other torque sensing BB's have their signaling.
I've recently written a general description of torque and PAS sensor signal levels here:
http://www.ebikes.ca/learn/pedal-assist.html#signal-levels
tomasinjapan
10 µW
- Joined
- Nov 20, 2011
- Messages
- 6
Hey Justin,
Sorry to come back to this old threat, but your experiences would be appreciated.
You mentioned:
"The earlier generation of Sempu sensors had the cable coming out the side of the device which was awesome as no modification to the bike frame was required, but those had other reliability issues. We're hoping these units are more rugged and first impressions seem to suggest as much."
What issues did you experience?
We are looking for high resolution crank sensor with eccentric BB (YST BB-2008-1C - http://www.yst-corp.com.tw/products-5.html). Sempu T2.2 would fit perfectly. Currently we use Piher PST-360 (12bit analogue absolute crank angle) with slightly modified BB. However, I doubt robustness and cost is high.
Ideally, we have an absolute crank angle measurement (either analog or digital, e.g. I2C), but I have not found anything affordable, robust and compatible with BB fitting except the PST-360.
I contacted Sempu to check if they do a position only sensor like Thun.
Any idea's appreciated. Thank you very much.
Regards,
Tomas
Sorry to come back to this old threat, but your experiences would be appreciated.
You mentioned:
"The earlier generation of Sempu sensors had the cable coming out the side of the device which was awesome as no modification to the bike frame was required, but those had other reliability issues. We're hoping these units are more rugged and first impressions seem to suggest as much."
What issues did you experience?
We are looking for high resolution crank sensor with eccentric BB (YST BB-2008-1C - http://www.yst-corp.com.tw/products-5.html). Sempu T2.2 would fit perfectly. Currently we use Piher PST-360 (12bit analogue absolute crank angle) with slightly modified BB. However, I doubt robustness and cost is high.
Ideally, we have an absolute crank angle measurement (either analog or digital, e.g. I2C), but I have not found anything affordable, robust and compatible with BB fitting except the PST-360.
I contacted Sempu to check if they do a position only sensor like Thun.
Any idea's appreciated. Thank you very much.
Regards,
Tomas
izeman
1 GW
Justin, would that mean that those SEMPU sensors would act like those old (Bionx?) drives that could sense the stress on the chain and start the motor just by putting your foot on the pedal?justin_le said:
Is this the way the CA interacts with the torque sensor these days? So will it start by stepping on the pedal IMMEDIATELY, or will it still need some revolution of the cranks?
And will this model https://www.ebikes.ca/shop/electric-bicycle-parts/torque-sensors/sempu-140.html be the correct one for my Kona Freeride with 83mm BB shell and a 150mm wide axle?
According to what I've been told in the CA3 beta thread, the CA itself requires 50RPM minimum on the cadence sensor input before it will start outputting a throttle signal.
(which is a problem for my setup that I'll have to overcome by creating a device that will output a pulse train to the cadence input until I can actually get to pedalling that fast, then shut off so actual cadence is supplied to the CA).
(which is a problem for my setup that I'll have to overcome by creating a device that will output a pulse train to the cadence input until I can actually get to pedalling that fast, then shut off so actual cadence is supplied to the CA).
izeman
1 GW
Thanks amberwolf. This is what i had in the back of my mind as well. I was hoping that this 'feature' was solved yet. This is not what i would expect from a torque sensing device (for PAS this is inevitable).
The only solution is to use a throttle to get the bike moving if it's too hard to pedal from a standstill.
The only solution is to use a throttle to get the bike moving if it's too hard to pedal from a standstill.
No, you can also do what I suggested above. There are also probably other potential solutions.izeman said:
izeman
1 GW
Then i may ask you to explain to me again what your solution would be. If you say: a device that adds pulses until you are able to pedal by yourself: what's the difference/advantage to/over a throttle?
And why not change the CA's algorithm to something that makes more sense, like 'ignore' PAS if speed = 0 and add throttle dependant on torque measured??
And why not change the CA's algorithm to something that makes more sense, like 'ignore' PAS if speed = 0 and add throttle dependant on torque measured??
I'd rather change the way the CA works to just allow torque-only PAS...but since I can't....
The advantage of using the device to make pulses to override that limitation means you don't ahve to use (or even have) a throttle; could control the bike entirely via the pedals. (I don't think it has analog braking options via the PAS, though, just on or off; can't recall).
In my case, the primary reason I want to use the PAS at all is because my hands sometiems just randomly go numb (and this gets worse as I get older). When this happens, I ahve to actually look at my hands to be sure I'm doing what I intend to, and can't always manage it. So using a throttle isn't always a reliable action; my thumb might slip off, or my grip be insufficient for a grip throttle.
(Im still working on braking solutions for when they're needed, but so far Iv'e still been able to squeeze a lever for regen braking easily enough, and mostly ok for the mechanical ones. Haven't had troulbe holding on to the bars ro anything, just changing what my hands are doing at the moment they go numb).
The advantage of using the device to make pulses to override that limitation means you don't ahve to use (or even have) a throttle; could control the bike entirely via the pedals. (I don't think it has analog braking options via the PAS, though, just on or off; can't recall).
In my case, the primary reason I want to use the PAS at all is because my hands sometiems just randomly go numb (and this gets worse as I get older). When this happens, I ahve to actually look at my hands to be sure I'm doing what I intend to, and can't always manage it. So using a throttle isn't always a reliable action; my thumb might slip off, or my grip be insufficient for a grip throttle.
(Im still working on braking solutions for when they're needed, but so far Iv'e still been able to squeeze a lever for regen braking easily enough, and mostly ok for the mechanical ones. Haven't had troulbe holding on to the bars ro anything, just changing what my hands are doing at the moment they go numb).
Hmm. I poked around to find the post about that, and it was here:
https://endless-sphere.com/forums/viewtopic.php?f=4&t=37964&p=1384020&hilit=50rpm#p1384020
but it doesn't actually say what I thought it did:
https://endless-sphere.com/forums/viewtopic.php?f=4&t=37964&p=1384020&hilit=50rpm#p1384020
but it doesn't actually say what I thought it did:
teklektik said:
amberwolf said:
Aha thats a different thing..
If i think about the "assumption of 50RPM as soon as it detects pedaling", it makes sense because human watts comes from RPM x Torque, and if the CA would not do this, the power at lets say 10RPM (human watts x multiplicator factor) would be very low and so the assistance of the motor.
I plan to swap out the Thun sensor for the Sempu in spring because the Sempu has 48 signals per revolution and together with the torque measurement on both pedals this should improve the response many many times, especiall during start where the Thun sensor now needs at least a few revolutions on the crank for smooth assistance.
If it doesn't work well enough, i will try out nucular 6F controller unit which directly supports the Sempu sensor in the controller (it is used on LMX 64 bike).
izeman
1 GW
AFAIK you can set min RPM to ZERO, but i can't find the settings right now.madin88 said:
The problem i have now: I can't distinguish one sensor type from the other. Is there a comparison what T2, T2.3, T4 etc ... means and what is used when?
https://www.aliexpress.com/store/product/torque-sensor-transducer-standard-buttom-bracket-parts-for-electric-scooter-e-bike-motor-assisted-bicycle/2802219_32793693999.html?spm=2114.12010612.8148356.2.51a21202RmGbpO
To just improve teh cadence detection you can use (or make) any number of magnets on the chainrings and sensor on the BB/frame, up to the max the CA allows (24? I forget).
But so far it sounds like the Sempu might be a better torque sensor than THUN or TDCM anyway.
I'd be willing to try one out if I could find someone that wants a used THUN for enough for me to get the Sempu. (THUN was off my old CrazyBike2--I never got to use it as a torque sensor there; using a TDCM on SB Cruiser right now).
But so far it sounds like the Sempu might be a better torque sensor than THUN or TDCM anyway.
I'd be willing to try one out if I could find someone that wants a used THUN for enough for me to get the Sempu. (THUN was off my old CrazyBike2--I never got to use it as a torque sensor there; using a TDCM on SB Cruiser right now).
Even if you can, it still wont' startup from torque alone--it does still require a minimum amount of pedal rotation (don't know how to tell/say what that is) before it will begin outputting a throttle signal based on the Torque/PAS input.izeman said:
If it *could* startup from torque alone it'd solve my problem, but I've been assured that it cannot.
izeman
1 GW
Can not by design? Or can not by state of software?amberwolf said:
My brain is a bit rusty already, as i have some conversation with teklektik in the back of my mind about this issue. But i can't remember what the outcome was. I will ask him to join this thread.
amberwolf said:
The CA applies power to the motor according to "human watts x assist factor"
At 0RPM or very low RPM the power would be close to zero so thats probably the reason why it assumes the 50RPM if it detects any movement.
The Sempu with it's 48 signals per rev should help the CA to detect any movement much quicker but it still will take some way on the crank.
stancecoke
100 kW
- Joined
- Aug 2, 2017
- Messages
- 1,811
justin_le said:
The Sempu 4.x measures torque, not bending.
You can find detailed information in the german forum.
older release
recent release
regards
stancecoke
Similar threads
- Question
