experience with THUN / x-cell sensors (and CA)

[izeman]

has anyone got some hands on experience with thun torque sensors?
i plan to install one in my 2.5kW middrive.
it's a NCTE isis compatible sensor, which is said to be 100% compatible to x-cell RT sensors.

any help warmly welcome!

 

[Triketech]

We have two Thun BB's connected to CA3's on recumbent trikes.

Both have been checked for load-voltage by hanging a weight from the pedal and measuring voltage. Both CA3's have the same settings. Both are installed with the same orientation. Yet one reads 50% higher Human Watt Values than the other. Its pretty easy to swap booms (and there fore Thun BB's) between the two trikes. I haven't figured out why such a deviation yet.

Despite the deviation either one works very well in pedelec mode. The CA3 is augmented with the trimpot for power multiplier, and one of the Thuns needs to be set a little higher on the trimpot to equal the actual output ratio of the other. No big deal, would never have known there would be that big a difference had we only bought one of them.

As far as how it works in pedelec mode, in a word, splendid.

With CA3 averaging set to 8 poles (cadence sensor) it takes about 1/2 turn of the crank for the motor to respond to inputs. Hub mounted torque sensors have to deal with a wide range of torque sensitivity as the force applied could be in 1st or 10th gear. In higher speed gears the resolution of a hub mounted sensor falls away. Not something you really notice until riding with a BB torque sensor. The BB Torque sensor is far better in my book.

 

[teklektik]

We have two Thun BB's connected to CA3's on recumbent trikes.

Both have been checked for load-voltage by hanging a weight from the pedal and measuring voltage. Both CA3's have the same settings. Both are installed with the same orientation. Yet one reads 50% higher Human Watt Values than the other.

... one of the Thuns needs to be set a little higher on the trimpot to equal the actual output ratio of the other.

Not to digress too far from izeman's topic, but:

• The cause of the variation is unclear, but since you have gone through the difficulty of applying test torque etc, you can always recalibrate one or both of the units by using the 'Custom' instead of 'Thun' torque setting.
  
  
  Here the default 2.5V offset will be corrected as per normal when you do the long Right Press. You need only adjust the scale factor so the two Thuns provide the proper Nm torque reading on the SETUP TRQ screen - or at least match each other. The Torque Scale factor maps the displayed voltage minus the Torque Offset to the displayed Nm, so you can jiggle the factor such that the torque reads appropriately even if there is a disparity in torque sensor voltage output. From there, proper (or at least consistent) HW and assist operation should follow...
  
  Not normally worth pursuing, but since you have two for comparison, it might be worth it so the HW readings match and the PAS Assist controls behave similarly.

 

[izeman]

thanks guys. teklektik your posts are never OT, and always welcome!!
i now found out (from justin's website) that THUN sensors only react on pressure on the LEFT PEDAL.
is this the same for the NCTE? this is know to be 100% compatible to the THUN X-Cell RT. but is it 100% identical on the mechanical part as well? that would be bad, as for me having to pedal on one side only to start the bike is a no-go for me.

 

[teklektik]

A quick email to Justin will probably get you the answer for that device - they evaluate a lot of different sensors.

At this time the CA3 Start/Stop logic is driven entirely from crank rotation, so asymmetric torque sensing has no impact on initiating assist. The configurable 'torque averaging' takes care of the 'missing' torque signal for half a revolution, so all is Good.

I don't like to get into 'futures' features and there is no implicit promise here, but since you are looking at a purchase and are concerned about this matter: There will likely be an upgrade in 3.1 that will give a snappier start to PAS based on torque. This should make the PAS system more responsive and more like the more sophisticated integrated PAS systems. This will operate in addition to the existing scheme so it's an optional performance enhancement.

Not sure how that affects your device purchase decision, but that's the 'all cards on the table' view from here...

 

[izeman]

thanks teklektik, i guess i now am even more unsure what to do
so you say, that the torque measured by the bb, which is translated into a voltage is NOT taken by the CA to start the bike? the CA is waiting for a PAS signal to start, and then takes the torque signal to decide to "how much" power it sends to the motor?

 

[Triketech]

Tek,

Yes, (and a thousand thanks for the guide). Also did some poking around with a scope.

Mine seems to be accurate. I also have a 48V (16S Li-Fe...)battery & MAC.

My wife's is a bit off only when running. At static the curve for a couple different weights lines up with the Thun graph. However her bike has a 16V (4S Li-Ion, Panasonic) and no motor. She uses it for human watts and that its lighted for night riding.

 

[izeman]

i just got an information from NCTE sales guys, that their BB is 100% the same as THUN as NCTE build it for them.
and they also confirmed that it only senses the LEFT crank.
this makes me unsure how it performs. i always though the system would work like those top-class haibike or similar, where you put your foot on the pedal, slightly put pressure on the pedal and the bike lifts off. the harder your press, the harder it accelerates.
if this is not what to expect with THUN and CA, then i guess i won't need it. if i have to do the hard part of starting the bike from a standstill and have NO motor support for the first half of crank revolution, than this seems needless. i thought i could lift the front wheel when hoping over an obsticale by just pushing the pedal hard.
this systems seems more aimed at "regualar pedelec" guys, that need a natural support of their motor while pedaling all the time. is this correct?
averaging the torque signal is working well for this use case for sure, but i guess not for mine?

could someone please confirm my thoughts?

 

[teklektik]

Yep - the CA3 algorithm was designed to fall in line with standard pedelec requirements that require actual pedaling to provide assist. Those 'torque-only' implementations clearly have applications and followers but this is not 'pedaling' in any real sense.

Torque sensing systems are essentially a basic PAS wheel with an extra torque sensing feature hung on the side. The CA presently initiates power application based entirely on the PAS wheel component. As you say, the torque component offers only the 'how much' part of the functionality. Unfortunately, this means the CA does not provide the kind of operation you are seeking.

 

[izeman]

 

[izeman]

i continued reading on torque bottom brackets, and guess the TDCM is maybe the better alternative for what i need.
in case there may be some change in the CA's signal handling, the TDCM will work with pressure on both cranks, even though it may not be as accurate as the THUN/NCTE sensors.
but as i'm not interested in precise results of my human power added to the system, i think this could be a great solution.
only real drawback is SQUARE TAPER norm. this system is soooo bloody old (ISIS/OCTALINK is long dead as well though) that you can't get any new downhill components anymore for this system.
maybe i still have some shimano xtr crank from the early 90s lying around.
are there any other CONS for the TDCM system?

 

[izeman]

has anyone heard of that one? https://www.alibaba.com/product-detail/ISIS-Torque-Sensor-bottom-bracket-for_60205374673.html?spm=a2700.7724838.30.10.6OurYm

 

[teklektik]

... the TDCM is maybe the better alternative for what i need.
in case there may be some change in the CA's signal handling, the TDCM will work with pressure on both cranks, even though it may not be as accurate as the THUN/NCTE sensors.
but as i'm not interested in precise results of my human power added to the system, i think this could be a great solution.
...
are there any other CONS for the TDCM system?

The TDCM has 12 poles compared to the 8 of the Thun and so can be adjusted with the present firmware to kick in a bit sooner - so a 'Pro'.

On a somewhat funny note: When discussing this left-side issue with J and the possibility of installing a TDCM in lieu of the Thun I have on hand, he observed that it wasn't difficult to get used to launching with the left pedal up - which was a "D'oh!" moment for me since I do that as a matter of course anyway! I stand on my right when stopped and just flip the crank around without thinking - I just never really thought about it in the PAS context. So - problem already solved....

 

[amberwolf]

Partly OT, but I thought Grin might want some input on this:

Those 'torque-only' implementations clearly have applications and followers but this is not 'pedaling' in any real sense.

Except...in cases of disability, such as with my knee pain (and other joints): On my heavy cargo vehicles, I can't even give a partial crank rotation from a stop on many days.


Even on a regular bicycle, I can sometimes only manage it without pain if I can lean against a building, rail, pole, etc., lift myself up by the handlebars (and hopefully not fall over), make my leg straight first, wihtout any weight on it (cuz bending it under a load is what hurts most), then put my foot on the pedal at about 1-o-clock and basically drop my weight down thru the straight leg to get started...once I'm moving, then as long as the bike has a low enough gear but is moving fast enough to not fall over, then I'll be able to continue pedalling along.


On a CeMoto upright bike I helped Cvin setup with CA3 and TDCM, I can only get it started if I use it's throttle first, and then use the PAS to continue. (or if I use the trick above that I use with non-motorized bikes).


If the CA would use the amount of torque to determine startup power, rather than rotation, I could use a torque sensor to get going and continue pedalling, too.

Right now, I would have to use a throttle to get started, and then pedal to continue. (presently I only have throttles setup on the bike and trike).


Now, in my case, I dont' usually pedal at all, partly because neither trike or bike is geared for anything but really low speeds (for the "in case all motors stop working" so I could at least try to pedal home if I'm capable of it at the time), and partly because it hurts even to ghost pedal, but if the CA could do this, it would encourage me to pedal on days it isn't so bad, and I could hope that either I'd get more used to the lesser pain or the exercise of the joints would help make it less to start with.


I"m sure it would make a difference to others like me, too.


And it would also make a difference to those with heavy cargo bikes and hills that they have to stop either at the bottom of before going up, or having to stop on the hill itself while going up (traffic controls, etc).

"We" can always add a throttle for those parts, at least in the USA, but for those that don't want a throttle for whatever reason, being able to use the torque sensor vs crank rotation to begin starting the motor would be helpful.


(except on CB2 I have a derailer to take up some slack in the long chainline and guide it from slapping around the frame...so if I could setup a cable to pull the derailer forward to let slack into the top of the chainline as I rotate pedals backwards, then I could rotate them forwards again to get that initial rotation into the CA to get it to begin applying power, and then continue from there. But I am not sure how well that would work, etiher...and it's just another "hack" to the bike.

Alternately, I could make some sort of dog-clutch in a jackshaft from pedals to wheel, and disengage it for initial startup rotation, then reengage for torque sensing and pedalling.


At the moment the sensor I have for the purpose, the THUN, only senses left side and that's my worst knee, so I'd have to set the assist multiplication higher than I would with the other one, I think.

(None of this matters ATM cuz my CA3 is still blown up from the CA-external-shunt's hall-speed-wire shorting to the B+ wire somehow in the cable from the shunt to it's CA connector; someday I'll get back to checking it out).

 

[izeman]

thanks amberwolf. i think you have a valid point here.

and i don't think that pedal pressure acitvated throttle feature is illegal anywhere around the world. eg haibike and other german companies build there bike that way imho. there are torque activated motors of several brands that you can buy. german law just says, that you have to pedal for the motor to work. and you do this in the moment you put your foot on the pedal.

what i always disliked with PAS was the delay. delay for starting the motor and for stopping it.
i had a long conversation with maddin yesterday about how to implement a delay-free torque based throttle. and it doesn't seem to be trivial. the control mechanism has to determine exactly WHEN you stop pedalling.
the way PAS works is, that a certain amount of time without PAS signal has to pass, before the controller stops the motor. this makes sense, as otherwise the motor would stop between any two PAS signals, making riding a very rocky pleasure but this delay, between stopping the pedalling, and end of motor support is sometime quite dangerous. so ebrakes are a must have.
but if you now know not only the rpm of the cranks, and if there is some pedal pressure, you must only apply power if there is some torque input (so NO ghost pedalling) and vice versa you can stop the motor in the very moment, this torque is no longer supplied.
does this make sense?

 

[Triketech]

Here the default 2.5V offset will be corrected as per normal when you do the long Right Press. You need only adjust the scale factor so the two Thuns provide the proper Nm torque reading on the SETUP TRQ screen - or at least match each other. The Torque Scale factor maps the displayed voltage minus the Torque Offset to the displayed Nm, so you can jiggle the factor such that the torque reads appropriately even if there is a disparity in torque sensor voltage output. From there, proper (or at least consistent) HW and assist operation should follow...

Not normally worth pursuing, but since you have two for comparison, it might be worth it so the HW readings match and the PAS Assist controls behave similarly.[/list]

Yes I have tried some similar things. Using a 15 and 45 Lb weight and calculating torque (based on her 150mm crank vrs my 160 mm cranks) and reading the values from the CA when the loads are applied with the crank at 3:00. In a static mode with Scale@-200nM/V and Offset@2.5V the static readings are pretty close.

I've also played with varying scales, offsets even increasing samples for averaging. I have a hunch that either the production tolerance consistency of the THUN is either wider than one would expect, or I bought a defective Thun. Would never have known it had we not had two, although the power output numbers do make my wife feel a bit proud.....

FWIW, I have a pretty lengthy background with AC Induction and BLDC system design; a current project is using a TI DRV8313 to drive a sensorless BLDC motor for zero cogging positioning at very low speeds, in a similar manner to how Applimotion zero cogging works.