No data - pickup sensor or barking up the wrong tree

[eglin73]

So for a bit of fun I have scoured the forums and interweb for information on how the concept2 data pickup works and the metrics are derived. I also looked at open rowing monitor, which looks great but decided to go with the tried and tested instead, and fit the PM5 to my dumb rower expecting there to be some issues with the metrics.

So having found info on the magnet arrangement in the forums, I adapted a ebike magnetic cadence sensor to fit on the hub and have 3 evenly spaced 3mm magnets ready to pass by the sensor. I thought it came out really great, and was pretty confident I could get signal. So I took the plunge and bought the pm5 and generator coil sensor, and setup a temp mount to hold the sensor and test the rig before making a permanent mount.

Only to find no data being transmitted at all. I expected the data accuracy to be the issue, I did not expect to get no data at all...

Unfortunately the board photo quota has been reached so I can't upload the photos but I just wondered whether anyone had any thoughts on why the magnets passing the sensor would provide no data?

Many thanks

Ian

[Citroen]

https://www.concept2.com/service/monito ... stuck-zero

Use https://imgur.com for photos.

[JaapvanE]

So for a bit of fun I have scoured the forums and interweb for information on how the concept2 data pickup works and the metrics are derived. I also looked at open rowing monitor, which looks great but decided to go with the tried and tested instead, and fit the PM5 to my dumb rower expecting there to be some issues with the metrics.

Disclaimer: as one of the lead developers of OpenRowingMonitor, I'm a bit biassed.

Fitting a PM5 to another brand machine is useless. Electrically, the construction of the PM5 and sensor is quite complex, and isn't understood by anyone aside C2 themselves. You need a very specific signal to trigger a PM5. Pulse length and intensity are quite specific on a model C and are hard to emulate. The generator coil sensor from a model D or RowErg is impossible to emulate as the required 12 pole magnet provides a dual frequency modulated sinoid which is extremely specific to Concept2.

Elecrical issues aside, several key parameters like flywheel inertia and sprocket diameter differ from machine to machine, making the output of a PM5 totally wrong. That is why the PM5 needs to know the type of Concept2 machine it is on. Even modifying flywheel or sprocket on a C2 kills the validity of the PM5 data as the underlying parameters in the PM5 are wrong. So, unless you have an original Concept2, the PM5 isn't going to give any valid results.

btw: we validated OpenRowingMonitor against a PM5 on a RowErg and a model D. Over 500 sessions (2000K in total) of testing shows that ORM is within 0.03% of the PM5 results, for each and every session. It is good enough for me.

[Carl Watts]

Ok here we go...

I learned quite a lot when I moved thorough each generation of monitor on the Concept essentially trying to fake a verified time.

The PM2 was pretty easy to fool, it was early tech and the pickup was simple and the monitor was slower.

The problem is however it will not work with a totally different flywheel because the C2 flywheel has a characteristic that the monitor is expecting to see. The PM2 had a much wider "Look up table" in regards to the flywheel, lets just say without being specific you could do things in a wider range and you would still get data and a reading .I'm not going to go into the specifics because I actually want all the results uploaded to the C2 website to be verified and I pretty much confirmed that unless the monitor was faulty or set wrong or both, its very hard to get fake result so getting one is defiantly beyond the average person to deliberately achieve.

Then you moved to the PM3 and this monitor got faster and smarter and suddenly I had problems even with the identical mechanical setup that gave you results with the PM2.Unfortunately many of the early PM3's developed corrosion in the Tach Op-Amp area due to the leaking memory backup capacitor and produced faulty results all by itself that were then easily verified.

The PM5 is next level again, it even knows whether or not the sensor is even connected or not, there is no chance its going to work without an actual C2 flywheel on the rower.

The Tach is not simply working on RPM's its far more complex, its working out the Drag Factor which is the base for producing a reading. The monitor is not only counting the pulses, its timing the pulses to work out acceleration and deceleration of the flywheel. From the testing and design that C2 has done what the PM5 now "Expects" to see falls within a very narrow band. Its complex based on the number of blades on the flywheel and the rotational mass of the flywheel. The PM5 is now fast enough to look at the pulses from the flywheel in more detail than the PM2, do what I believe is a more precision calculation because its faster and has time to do it rather than just a "Look up table" and only expects to see the tach pulse within a very narrow range that matches the known characteristic of the flywheel. If the PM5 doesn't "See" very close to what is expected, not just what is coming into it and processing that regardless, it produces no reading at all.

The bottom line is that no other flywheel is going to work. You simply cannot retrofit 3 magnets to another manufacturers rower and get it to work.

You have to give it to Concept 2, their rower is simply the best when it comes to the IP in the monitor. You can copy the hardware easily enough but all the smarts and what you are paying the extra for is in the monitor.

[JaapvanE]

The Tach is not simply working on RPM's its far more complex, its working out the Drag Factor which is the base for producing a reading. The monitor is not only counting the pulses, its timing the pulses to work out acceleration and deceleration of the flywheel.

I think it is much smarter than that, to be honest. OpenRowingMonitor behaves as you describe: we measure time between the signal reaching a certain threshold (12V), convert that into old-fashioned pulses and then do the math.

In theory, C2 could do the same thing inside the PM5. But C2 seems to have deliberatly added a second frequency shift over the signal per rotation (we sampled several machines and even spare flywheels sent from C2 to us had this). This frustrates ORM, but that is not my point and I don't think it is C2's either. This type of frequency modulated signal is a known approach for analog tachometers where the speed is determined by the "compression" of the sinewave, theoretically allowing infinite sampling (as it is analog). Getting the magets configured to make this work is extremely precise work. For some calculations, like drag or distance, having such an analog approach doesn't have any influence. However, for more advanced metrics like force curves, it is quite interesting.

From the testing and design that C2 has done what the PM5 now "Expects" to see falls within a very narrow band. Its complex based on the number of blades on the flywheel and the rotational mass of the flywheel. The PM5 is now fast enough to look at the pulses from the flywheel in more detail than the PM2, do what I believe is a more precision calculation because its faster and has time to do it rather than just a "Look up table" and only expects to see the tach pulse within a very narrow range that matches the known characteristic of the flywheel. If the PM5 doesn't "See" very close to what is expected, not just what is coming into it and processing that regardless, it produces no reading at all.

That is typically captured in the "flywheel inertia" parameter, which is the basis of almost every calculation in a rower. C2 knows their flywheel and indeed expect some behaviour and outside that the PM5 propably would assume something is wrong.

I don't know if they actually filter for outliers. In comparing ORM and the PM5 (we structurally compare to the golden standard as a validation of our own physics model), I noticed that the PM5 typically reacts to the start about 0.5 seconds slower. In the bluetooth specification, C2 indicates that there is a state "waiting for minimum speed", suggesting some filtering. But we can not replicate startup behaviour from C2, regardless of our settings (we can specify a minimum flywheel speed....), suggesting that C2 has a few tricks up its sleeve we mortals don't understand.

The bottom line is that no other flywheel is going to work. You simply cannot retrofit 3 magnets to another manufacturers rower and get it to work.

Agreed. A PM5 is a very purposebuilt monitor, and providing it with something outside its normal signal will frustrate it. OpenRowingMonitor can handle this, but requires the owner to do some calibration (and thus measure/estimate parameters), and even then some machines aren't suited as their build quility is too poor.

You have to give it to Concept 2, their rower is simply the best when it comes to the IP in the monitor. You can copy the hardware easily enough but all the smarts and what you are paying the extra for is in the monitor.

I like to think you pay a bit extra for the decent monitor, but also for great built quality, servicability and great service. My NordicTrack had a seriously unbalanced flywheel, and they refused to service it (despite it starting after a repair) until physical damage would be visible. I never had such a discussion with C2.

[eglin73]

Thanks for the feedback Jaap, Citroen and Jaap, even though it seems like I'm barking up the wrong tree.

I guess ORM might be my best option, and at least not all of my efforts are wasted in the hardware side and it's been a bit of fun learning about both the PM% and the ORM.

Thanks again for taking the time to answer with so much detail - much appreciated

Regards
Ian

[eglin73]

Just in case anyone is interested in seeing my handiwork, I uploaded the images to imgur as Citroen suggested

https://imgur.com/gallery/tDGxZGZ

Regards

[Tsnor]

...

Amazing content. Thank you. (Wish site had +1 counters to stroke). This discussion of how PM5 works between you and Carl is pure gold. It doesn't feel like sticky information, but this is likely the only thread on the planet where you can learn this.

[Carl Watts]

Well you never stop learning either.

Try this one for no pickup data that I fixed just last week.

This is the worst PM2 I have ever seen but managed to get it going again.



Got it back to the original fault and probably the reason it was stored in a shed without the battery cover so it turned into a bit of a hive for masonry bees that literally drag in dirt and sand to built their nests.

Replaced the usual suspect Op-Amp but for once it didn't fix it. Discovered the output of one Op-Amp goes through half of the comparator also used by the Heart Rate input and had never had one of those fail in over 10 years.

Cleaned up pretty well, fortunately they built in the area reserved for the PM2+ only. Clear coated the entire board so even if they built there again it would keep running.

[c2jonw]

Well you never stop learning either.

Try this one for no pickup data that I fixed just last week.

This is the worst PM2 I have ever seen but managed to get it going again.

Just Wow!

[Ombrax]

This is the worst PM2 I have ever seen but managed to get it going again.

Wow, Carl - you must really love a challenge!

Edit: I see Jon and posted nearly the same thing!

[JaapvanE]

Well you never stop learning either.

Try this one for no pickup data that I fixed just last week.

This is the worst PM2 I have ever seen but managed to get it going again.

Wow! I would have assumed it would be a terminal case, impressive.

[Carl Watts]

Cheers guys, thought it was dead but soaked it in a bucket of hot water for a bit and brushed off all the muck and hit it with compressed air.

It powered up okay, just didn't show the tach input as working.

Battery acid leaks are the thing that really kill these, I recommend anyone running a PM2, PM2+ or in particular the DYNO monitor only ever run Lithium AA batteries.

[jamesg]

Only to find no data being transmitted at all. I expected the data accuracy to be the issue, I did not expect to get no data at all...

Compliments to C2. Their machines can't be hacked or copied. Thanks for confirming this.

Just noticed April 1st went by without comment.

But maybe it was this:
Workout of the Day: 10 x 2:30 / 30 seconds easy