Throttle-In variations with JLog 2.6, Taranis, and Castle Edge ESC

[RedOctobyr]

Tom, I recently bought a JLog 2.6 GW, and got it configured and wired into my Castle Edge 100. I'm using a Taranis.

Setting it up went OK, piecing together info with Google Translate, and I made my own "Y" cable. I'm using firmware "26-CC_FR-128.3.bin", and configured it with JLC7. I'm using Castle filtering mode 2. I was pleasantly surprised when it suddenly just worked Very nice.

I got in 3 flights in today. It went very well overall, but I have a few concerns. I am seeing some odd motor behavior, which is making me a bit nervous. It seems as though the throttle signal going to the ESC is changing slightly at random.

I am using Autorotation Bailout on my Edge, and Set RPM governor mode (so it only looks for 30%/70%100% throttle signals, to select 3 different headspeeds). For Autototation Bailout, that requires that Throttle Hold is set to a certain number higher than 0% (or -100%, on a Taranis). It's just low enough to shut the motor off.

After the ESC arms, while in Throttle Hold, I will get twitches from the motor. It will suddenly move for just an instant, maybe half a motor revolution, etc. I have Throttle Hold set to -68%, to make Autorotation Bailout work. If I lower Throttle Hold to -100%, these twitches go away. I tried -72% (you can't go too low, or Autorotation Bailout won't work), and the twitches continued.

In addition, during flight, I got what I could best describe as "stutters" from the motor. It wouldn't shut down (thankfully), but I'd hear the motor's sound suddenly change, just for an instant. This was at the middle governed RPM, the 70%-throttle speed. All I can assume is that the ESC briefly saw a higher or lower throttle signal, and started to change RPM.

When setting up the JLog, I reduced my ESC logging speed to 1Hz, from my usual 2Hz. But I looked at a 2Hz log, from before adding the JLog, the Throttle In line is perfectly flat. But even looking at the 1Hz log from today, running the JLog, I can see variations in the Throttle In line.

For Throttle Hold, the average is the same 1.190ms as I was getting before, but with variations. One dip that was captured during flight went from 1.668ms to 1.621ms. Another went from 1.903ms to 1.848ms.

Some screenshots of the Castle logs are attached. The only line shown is Throttle In. Ignore the large jumps in throttle before the main portions of the flights, I was going in/out of Throttle Hold, checking Autorotation Bailout. I have raised the ESC logging rate, to see what it captures.

Do you have any ideas/suggestions? Thank you.

[RedOctobyr]

I've done some more testing and logging with the Edge ESC, to try and better understand the problem.

I logged some flights at 10Hz on the Edge, to try and capture the glitches.

As a note, the Taranis uses a -100 to +100 range for values, not 0 to 100. I'll use the numbers shown on the TX. So divide the differences by 2 if comparing to a Spektrum throttle curve, or the way Castle expresses their numbers.

Things I've learned:
- Changing my throttle value by 1 on the TX corresponds to a 0.008msec change in the ESC logs. Giving me a way to help correlate between the two.
- The motor starts to spool at -66 on Throttle Hold. I've been setting Throttle Hold to -72 (following the Autorotation Bailout setup instructions), and still getting motor twitches. I was using -68 without the JLog, and never had any motor twitches. So the signal is sometimes going to, or higher than, -66. That's a change of at least +6 on this scale, or 0.048msec.
- The ESC logs show spikes and dips as large as 0.047msec in-flight. Corresponding to almost +6 and -6 on the TX's scale.

So far this is mainly an annoyance (though it makes me nervous when the motor sound suddenly changes while flying). And of course motor twitches while in Throttle Hold are bad. But a concern would be going into Throttle Hold for autorotation practice, and having the signal dip during that time. That could trigger the ESC to go back to a soft start, instead of the quick spoolup. So when I'd come out of TH near the ground, it might slowly speed up again, causing a crash.

I really would like to get this resolved, and would appreciate any insight or suggestions you could offer. I believe I have it all wired up correctly. I will try to do any tests you suggest. As this is the 2.6 GW, I can't offer any logs from the JLog itself (if those would help), merely from the ESC, and what the JLog sends back to the Taranis. A screenshot of one 10Hz Castle log is attached, along with the actual .csv log file.

If I should make a new thread for this, just let me know. Thank you.

[dl7uae]

Well... as I already wrote several times in the two support forums: I'm struggling with the simple throttle pulse inversion function by JLog already for a long time. Theoretically little twitches are possible because pulse inversion is based on an interrupt routine and the processor used in JLog does not provide interrupt prioritization. And we have a huge number of parallel processes triggered by interrupts.. The reaction time of the interrupt-driven pulse inversion routine (see a pulse polarity change in input, put the pulse out inverted) is some kind of non-deterministic so to speak.

On the other hand: In all cases I had the chance to test with user's equipment by myself it turned out that the problem was caused by cabling, by a ground loop. I replaced the cable salad from the user by our "Y harness" and odds gone.

...

[RedOctobyr]

Thank you for the reply. Let's start with the easy things, wiring I built my "Y" cable following your diagram of the "Y+" cable. But I will check it again, perhaps I wired it incorrectly. Unfortunately, the vendor I bought the JLog from does not sell the Y+ cable, which is part of why I made my own.

I followed the "Y+" diagram (26-CC_FR-3) in part because I found it the clearest to understand, with the best labeling for each wire, showing exactly where it goes.

The "Build your own special Y" diagram (26-CC_FR-2) was slightly less clear, for me. And I will confess that I'm not sure what the basic connection diagram (26-CC_FR-) is really trying to show

I am not familiar with ground loops or what I may have done to cause one, but I will do some reading.

I am using an external BEC (Castle BEC Pro), at 6V. The red wire coming from my Edge 100 ESC (not an HV-model) has been removed from the ESC's servo plug. I removed the red wire from the telemetry cable that goes from the JLog to my other telemetry sensor, so it should not be getting voltage that way.

The throttle signal goes from my X8R RX into my iKon (using SBus), then goes from the iKon into JLog. I can take it directly from the RX to the JLog, if that would be better.

[RedOctobyr]

To start, if this is simply unrelated to using a Taranis, please feel free to have a moderator move these posts to a new thread. I honestly don't mean to clutter up this thread. I just wasn't sure if my issue was specifically Taranis-related.

I have taken a few steps to simplify my wiring, and try and "consolidate" my grounds.

- I removed the ground wire from the FrSky telemetry cable going to the JLog.
- I went as far as using a single lead from the BEC Pro (normally I use both of its leads) to a Y-cable off an iKon port, to power the system.
- I left my Scorpion BUG turned off.

Some of the changes were tested with flights. The later changes were tested indoors, by letting the ESC arm, and keeping the heli in Throttle Hold for a few minutes, then checking the ESC log.

During those Throttle Hold tests, I still heard twitches from the motor, and the log still showed Throttle In variations. The next time I can fly it, I will see if the variations look any better than they were originally.

I will look through my wiring again to see if I can find any issues, or something else that might cause a problem with grounding.

[dl7uae]

Hi!

I do not have a mod role here.

Will test this behavior by myself on the bench.

[RedOctobyr]

Thank you Tom.

One thing I realized this morning is that a few weeks ago I started using the Edge ESC's white Aux wire to control the governor gain. In the logs, I noticed that the governor value was not smooth, it would sometimes jump up and down. The Throttle In line was still smooth. This was all before adding the JLog.

So something else on the ESC logging also shows variations, though I don't know if that has any significance. I will try disconnecting the Aux wire (which is just a signal wire, no ground, etc), just to see if that changes anything.

I just checked a log, the governor gain and the Throttle In lines jump at different times. I was curious if the variations would happen at the same time; they do not.

[RedOctobyr]

Originally Posted by dl7uae

Hi!

I do not have a mod role here.

Will test this behavior by myself on the bench.

Tom, have you had a chance to look at this?

I unplugged the Edge ESC's Aux wire, to further simplify the wiring. All my grounds go to the FBL, and go out to the different devices from there. The ESC just has its main lead plugged in, with the red wire removed (I'm using an external BEC).

I am still getting the Throttle-In variations in the logs, including audible changes in the air, and motor twitches while in Throttle Hold.

I don't know if temporarily loading a different firmware might show anything? I could load one for a setup that physically connects the same way for the Castle ESC wiring, and temporarily give up JLog telemetry, for just testing it on the ground, with the ESC armed. I don't know if that would help with understanding things, if the FrSky-specific FW could influence something like this.

Thank you.

[dl7uae]

Tom, have you had a chance to look at this?

Not yet, sorry!
I am "currently" in two m-o-n-s-t-r-o-u-s (Huh?! The forum software changes the word..) developments..

Will do this within the next three days, promised.

[RedOctobyr]

I'm sorry, I didn't mean to sound like I was rushing you.

I will be traveling for about a week anyhow, so I won't be able to do/check anything with the heli itself.

[RedOctobyr]

I know you must have a lot going on, but have you been able to look at this? Thanks.

[dl7uae]

Everything built up to the test. Of course, the battery of the transmitter is empty. Just charging..

[dl7uae]

Well .. nothing, no unusualness.





Idle:
Some "artefacts" via CC data link regarding throttle "echo" as well as on other data but absolutely no misunderstanding of the throttle pulse by the CC.

Run:
Everything ok, throttle "echo" w/ at max 2% of variation, no misinterpretation (by CC) or deviation in echo, no jumping in pwm nor rpm..

[RedOctobyr]

Hmm, OK. Thank you for doing the test, I appreciate it. I will continue looking into things on my side.

I will double-check the cable that I made trying to follow the Y+ cable diagram, as well as trying a few other things.

I looked at the first part (IU1) of one of my 10Hz logs in Excel. Compared to the average of that segment (1,6677 usec), I got spikes up to +2.82% vs the average, and as low as -2.82% vs the average (47 microseconds in each direction, up to 1,7147, down to 1,6207). The spikes vary in height, but the largest spikes were the same amplitude up and down, which seems odd.

For curiosity, what generated the black-background graphs? In "Run", your ~1,000 microseconds red line (Throttle Pulse Length microseconds) seems to show some ~50 microsecond variations, despite them not appearing in the Castle graph viewer. Those are similar magnitude to mine, though mine *do* appear in the Castle graph, and cause motor twitches in Throttle Hold and during flight. Am I reading this correctly?

Thank you for looking into this. Your motor test setup looks pretty cool!

[dl7uae]

For curiosity, what generated the black-background graphs?

In "Run", your ~1,000 microseconds red line (Throttle Pulse Length microseconds) seems to show some ~50 microsecond variations, despite them not appearing in the Castle graph viewer. Those are similar magnitude to mine, though mine *do* appear in the Castle graph, and cause motor twitches in Throttle Hold and during flight. Am I reading this correctly?

It seems not to affect the throttle understanding of the ESC.
Guess this is only on the "throttle echo" (data output) from the CC.

[Vinger]

Can we please see some more photos of your test rig, looks awesome, maybe put it in a new thread. Curious about the motor coupling as well as what you use for motor brake of slave motor to determine power consumed?

Really jealous.......

[RedOctobyr]

Guys, I asked the mods to move the posts on my signal issue from the Telemetry #8 released: FrSky S.Port thread to a different thread. I didn't want to keep cluttering up the "main" FrSky S.Port thread with an issue that may be specific to something with my particular setup.

I have started doing some testing with a digital logic analyzer. It's an 8-channel USB unit, up to 24MHz sampling, $12 on eBay, heavily-inspired by the Saleae Logic unit. It's my first exposure to something like this (I'd never heard of logic analyzers until doing some reading recently), but it seems like a really cool tool. I can finally see the PWM signals that run everything on our machines, you can see refresh rate, pulse width, etc, for up to 8 signals at once.

I have started looking at the different signals, going into the JLog, and coming out of it, to try and understand where the variation appears (eg- is it just coming from my RX?), and how big the variation is.

[dl7uae]

Hey, good idea!

I did that too in the past. It is somewhat complicated to evaluate.
Reason was that there can be in fact some kind of variation. In principle it is a simple function to read the throttle pulse and put em out inverted. Problem is that there are many interrupt-driven procedures in parallel in JLog and that its microcontroller has no interrupt priority system. Some protocols are very time sensitive, you cannot allow to get interrupted in between, s.bus2 especially.

A Saleae-like logic analyzer for $12? Wow! I like the Saleae. It is not very expensive, sufficient for many applications and less "inconvenient" compared to other LAs - as DigiView to name one.

[RedOctobyr]

I did some testing, I found the results to be interesting.

I recorded about 300 seconds of data at 1MHz with the logic analyzer. The 1MHz sample rate implies, to me, that I should be able to measure pulse widths with 1-2 microsecond accuracy/consistency.

The signal in to the JLog, from the RX, is very stable. During the long portion of the log where I just stayed in Throttle Hold, it measured 1,151 microseconds, with a min of 1,151, and a max of 1,152. The error range was +/- 0.04% from the average.

The signal *out* from the JLog to the ESC varied a lot. It measured an average of 1,151 microseconds, with a min of 1,094, and a max of 1,206. The error range was +55 microseconds, -57 microseconds. Or +5.5%, -5.7%, if we assume a total signal range of 1000 microseconds. For a total error range of 11.2%. I ignored the "extra" pulses back from the ESC (sending data) in these calculations.

Graphs of the two signals are attached, along with a graph zooming in on the output from the JLog.

In post #13, one of your Logview pictures showed larger spikes & dips in the red throttle signal, similar to mine. These did not appear in your Castle log, but they definitely show up in mine.

In my opinion, a +/-5% throttle signal range does not seem normal. I'm definitely not an expert, I admit But it seems like a lot to me.

I am happy to post/send the Saleae logic data capture file if that's helpful at all, or whatever info you would like. This test was done on the bench, so the motor was off, etc.

I do notice that there seems to be something of a pattern to the spikes and dips. The clusters of variations are somewhat regularly spaced, maybe ~80 seconds apart. For what it's worth, when I would find an abnormal number in the log, it was typically just one. Like the signal changed for just a single pulse, then went back to normal, if that means anything to you.

I appreciate your time and help with this. I am honestly not trying to just be a pain. I could have removed the JLog and gone to a FrSky sensor. But I like what the JLog offers. Rather than just give up, I'm hoping to be able to help learn something about the issue, and possibly even contribute (somehow) to improving it. I appear to be the only FrSky + Castle + JLog user who has posted so far, so at the moment, I guess I am your test site

Would it be helpful for me to load a different configuration to the JLog? Maybe telling it I'm using something common & proven (Spektrum, Futaba, whatever) with a Castle, and bench-testing the signal that way, to see if anything changes?

[dl7uae]

I will come back on this later. I'm on the roads.. just eaten 550km, now have an appointment, then again 270km..

Perhaps this evening..