START HERE |
|
Register | FAQ | PM | Events | Groups | Blogs | Calendar | Mark Forums Read |
Unregistered
|
JLog - The JIVE logging stamp JLog - The JIVE logging stamp - Official English Support Forum |
|
LinkBack | Thread Tools | Display Modes |
12-01-2014, 07:23 PM | #1 (permalink) |
Registered Users
Join Date: Feb 2008
|
JIVE Pro
Just to recall it: Why do we speak here about that all? Because it affects JLog and CVS16 (and other connected electronics). It creates support effort. Crazy BEC: It managed it also several times to drive a serial EEPROM in CVS to madness, calibration data have been deleted. (Hopefully new CVS firmware will be more resistant against that..)
Data Protocol Had a 120 Pro on the logic analyzer (Pro --> "TelMe" JLog): Yes, it's absolutely clear, there stepped a bug into Pro's firmware: - The value for the BEC voltage is frozen at 0.9 and comes up at a time after powerup at which all other values have not been updated yet. - Imot (motor current) gets frozen at the last value if commutation stops. ---------------- BEC Took the time now, not much, for a 1st view only.. We bought a 120 Pro 8P. Unfortunately.., looks like we got a "good" one. No time / no desire to analysis.. Just showing scopes.. In a nutshell: The BEC is based on a synchronous buck converter, the KOSMIK has a TI LM5116, in a JIVE Pro we presumably have it too. (This is the same chip as in the old JIVE HV, in the LV the LM3075 was used which can only 36V input, the 5116 can withstand up to 100V. - He may, but is unlikely be operated free running.) The current limitation of the synchronous converter is likely cycle-by-cycle current limiting. The cycle time is something around 1..4us (up to 1MHz) - but we know that the switching frequency is about 250kHz, so 4us cycle time. Thus, the current limit is very responsive. The chip has a passive circuit, inter alia, a resistor cascade controlled by the processor (STM32) to adjust the output voltage. The whole thing gets input voltage, starts, but the STM needs some time to initialize. - The behavior of our copy is very similar to KOSMIK. It is unacceptable that -- a) the voltage rises in such a slow ramp with plateau(!) phase * 1) * 2) -- b) slightly load at powerup folds the ramp down to about 2 volts. * 1) So the software of the OptiPower UltraGuard got a "special treatment" for the KOSMIK from us. The UG measures the R/C voltage when it comes up to determine his step-in voltage (minus 0.5V). * 2) Although the BEC of the old JIVE used the same controller - behind the BEC was a FET switch (later) activated by the steam-operated ATmega 168. - The anti-spark of the Pro is not only less effective than that of the KOSMIK, - remaining sparks also affecting the BEC voltage - in contrast to the KOSMIK. I used only 25 volts input voltage (6S)... My feeling is there, somewhere it lacks of capacitors or the right ones.The smallest wiggle when plugging brings electronics a brownout. (not observed with KOSMIK). - While the voltage at the option ports of the KOSMIK is stabilized 5.0V - comes just the BEC voltage out at the option port of the Pro. This load-dependence of the powerup voltage rise can cause effects in any conceivable variety. Actually, you can only advise to have separated all electronic load during powerup. That's not all just stress regarding the valid throttle pulse to get Ubec ever to setpoint. It could happen anyway that the ESC itself is concerned, finally his electronics is supplied by the same voltage.
__________________
Tom (j-log.eu) Last edited by dl7uae; 12-02-2014 at 04:28 AM.. |
Sponsored Links | |||
Advertisement |
|
12-02-2014, 05:39 AM | #2 (permalink) |
Registered Users
Join Date: Oct 2006
|
Thanks for sharing this Tom.
Looks about the same as what I measured, your scope shows some more detail that mine misses. BEC rampup to 6V (with 4 servo's) that I measured on the J120Pro: So basically the same as yours, minimum voltage in my case is ~3V. Also, what voltage do you consider dangerous during startup, is that 2V? Btw what load are you using for your tests? When I hook up 4 servo's during startup I typically get this with the J120Pro: So the rampup to 6V is followed by ~0.7sec of flat 6V voltage where the servo's kick in. Remarkable are the dips generated there: minimum voltage 4.5V. This was without blades on the heli so when the servo's have to work right at the start it migh get worse.
__________________
Raptor E700: PowerDekker Pyro650, JR DS8925HV & SPG01, 12s, Vortex VX1, Kosmik160, Jeti DS-16 Last edited by _Dekker; 12-05-2014 at 12:41 AM.. |
12-02-2014, 05:55 AM | #3 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
What is the difference here between Pro and KOSMIK except that Ubec comes out of the Pro option port instead of fix 5.0V (KOSMIK)?
The difference is the size of the input capacitors (Low ESR capacitors). In KOSMIK 5x 390uF , the 120 Pro only 2x 330uF. Our aim is not to simply operate only finger pointing, we want to present a solution.. * - A solution was already announced, the Protector with its added function. (That would be virtually the replacement for the FET switch behind the BEC in the old JIVE.) * - I must still test/measure but we believe that an external anti-spark resistor could cure it. Thus: To be continued.
__________________
Tom (j-log.eu) Last edited by dl7uae; 12-06-2014 at 11:49 AM.. |
12-02-2014, 06:10 AM | #4 (permalink) | |
Registered Users
Thread Starter
Join Date: Feb 2008
|
Quote:
(Did it at home, not in the lab at Zurich. There we use adjustable electronic loads of course.) The world of loads is colorful. I have therefore chosen a load that may be somewhat representative of problems, especially as our Pro obviously is not a "problem-Pro". I already had multiple cases (users, Pro's + setup==load) in which the BEC voltage never went above 2 volts. The CbC current limiting seemed to prevent that the voltage can get higher. But I never got such a Pro in my hands. K replaced these JIVEs then.
__________________
Tom (j-log.eu) |
|
12-02-2014, 02:46 PM | #5 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
An external anti spark resistance does not change much, so a series resistor, however, was quite enlightening..
The question is, who plays during startup at the voltage, the built-in anti spark stage or the BEC itself. The expense is already borderline.. Possibly we dissect a KOSMIK to determine the cause. However: What brings us (R2) this? Nothing, except to can say the user where a design flaw is in both ESCs. That's not our mission. Since we're talking about a foreign product, with the aim to offer an user feasible workaround. Our tests with external anti spark resistors (series resistors to exclude the supposed current limit) as well as a precaution also with fat low ESR electrolytic capacitors at the input and output - but now tells us already: It can not come from us, what it cures and at the same time would by user executable. We can only determine whether the anti spark FET circuit or the BEC itself is guilty. Let's see, maybe curiosity wins about rationality.. The remaining question is: Who builds the "kinks" in the voltage curve, the input (anti spark stage) or the BEC itself: This is still an "interesting" (and for R2 pretty useless) "prize question" between Linus and me. I currently bet on the anti spark FET circuit, Linus keeps both options open.
__________________
Tom (j-log.eu) Last edited by dl7uae; 12-02-2014 at 04:38 PM.. |
12-04-2014, 10:45 PM | #6 (permalink) |
CHC Benefactor
Join Date: Feb 2006
Location: Singapore
|
Very helpful analysis. I always start up with buffer pack on BEC line before powering Kosmik except on bench. The interaction with some FBL was always unpredictable. This was always blamed on additional load created by servo's and satellite receivers on startup. I suspected that was just a compounding factor which clearly your work seems to indicate.
|
12-05-2014, 05:33 AM | #7 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
Well, if we now investigate in the real inner cause or not.., one thing is clear already: there cannot be a workaround advised from us, pity.
To avoid any misunderstanding by readers who are new to the topic - let me summarize: The BEC in KOSMIK and in JIVE Pro shows at power-up a "quirky behavior" of its output voltage - a "rampup with kinks and dips". Why it only with the Pro came so much into effect for some users, we can not say currently. The assumption is that the underlying design flaw gets some "tailwind" by copy variation, or it is by the fact that a Pro for reasons of space has smaller low ESR capacitors. 1. The "voltage ramp with kinks" may interfere powered electronics or this does not happen. Reason: - It is also JIVE Pro copy dependent. Although K claims it affected only a few devices of the series JIVE Pro 80... but that seems in light of other experiences of users, including those of the 120 Pro, so as not to be correct. - It depends on the setup in the model, namely the nature of load at the BEC. There is no flat description because particularly dynamic aspects play a role: impedance curve of a powered device which is a load to the BEC. "Curve" because the BEC supplies a voltage rampup. Kinks and reversal points in the ramp leading commonly to a more non-linear impedance behavior of powered devices. 2. If something happens then only during power-up (battery to the ESC). After one second it's all over. Particular problems that occur afterwards during operation, i.e. during the flight, are not known! - In appropriate circumstances (ESC copy + load components at his BEC) it may happen that the whole thing never comes through the power-up phase, - i.e. the BEC voltage may never achieve its final value in stable condition. It is then clearly noticeable, - there is no risk of accidentally taking off in such a condition. Again and more precisely: Those who do not immediately after power-up experience problems with powered electronics (with a FBL for example) can quite relaxed rely on the BEC. Due to the design, however, it is advisable to offer the BEC not too much "backward voltage" from servos. A buffer can help a lot. A R2 buffer can convey the power-up problem of the BEC despite adjustable charge current limit, although it then (after power-up) would be the ideal solution in regular operation. A R2 Anti-Reversal Circuit (as option to the buffer or standalone usable) keeps reverse voltages away from the BEC but has no other buffer effect. A "dumb battery" would be an option but just uncomfortable in terms of battery management. An Optipower UltraGuard is good as a backup in case of BEC failure but hardly effective as a filter against reverse voltages. - The future "Protector" lasts all voltage madness from a BEC away from load, also allows the use of the R2 buffer without a "side effect" by the ESC/BEC bug. -------------------- If a user is having problems, he has 3 options: 1. Disconnect every load from the BEC, then ESC power-up, then reconnect. 2. Buffer voltage source at the BEC output: switch on or connect. ESC power-up, then buffer switched off or disconnected or let it on for general safety reasons.. 3. To wait a bit until the Protector appears in shops. Comes between (any) BEC and load. As a goodie on top it also keeps such voltage rampup away from powered devices.
__________________
Tom (j-log.eu) |
12-08-2014, 08:34 AM | #8 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
Please find attached the final, - examines the KOSMIK because non-destructive, - the principle is the same for JIVE Pro, the effects also.
No idea why it more often comes into effect for users with the JIVE Pro. Perhaps it is due to the lower capacity of the low ESR caps (Pro 120: 2x 330uF, KOSMIK 200: 5x 390uF), I'm completely unsure. Completely unsure: The ambition is not to make the design flaw circuitally identified. In this context, once again stressed: This 2nd act is purely "sporting nature". As a result, there are no recommendations for a better workaround than the 3 options that have been mentioned above. The examinee The voltage at the low ESR caps is measured (behind the anti-spark circuit), which is the single wire - also was inductively coupled to the stray field of the inductor of the BEC in order to be able to see what the synchronous buck converter really does. Anti-Spark Circuit (On the other side of the PCB is nothing to see except of a cap at the gate of the anti-spark FET.) (Pics "stolen" from here. ) Without having now taken up the circuit completely: It's obviously just a FET (one). It switches the low side of the caps. No wonder that the KOSMIK (as certainly the JIVE Pro) is not protected against reverse polarity. Too little challenging for a 800 dollar stone.. ----------------------------------------------- Outside of the protocol: few mechanical-electrical details of KOSMIK makes my hair stand on end.. * 1. The screws holding the power board is bolted to the aluminum plate are too long (or the hole in the aluminum plate is too short). The heads of the screws are out so far that it is critical with the distance from conductive points on the board above - bottom side, containing the 3 double-drivers for the power stages. * 2. No idea why the 5 low ESR capacitors are soldered 1.5mm distant from the board. Despite of intercepting the tops of the cups by foam material that may not be good in terms of vibration affecting. It could also very well be that they deliberately chose the screws too long. At the 5 terminal lugs the distance of the Direct FETs to the aluminum plate is made. The screws in the middle of the PCB could bring this too close to the plate when it is tighten firmly. The insulation is anyway borderline.. * 3. The three screws of the terminal lugs (motor side) are too long. They hit the three SMD connectors for the option ports and cause an inter-board connector not fully engaged in its socket on the power board (ADC <-> sensors). The MCU board is thus not completely parallel to the other, but anyway it is on the frame on when fully assembled. If the whole thing vibrates now ... Is that why the center PCB (BEC at top, drivers at bottom) of the triple sandwich is glued to the plastic frame? Or is that based on a constructive misunderstanding as to the PCB pads in the plastic frame? Well..., and to all the housing construction without shielding effect one must probably nothing say .. Plastic midsection, aluminum parts are not connected to ground. In addition: The thin anodized coating of the aluminum plate plus thermal grease also seems to me not to be a sufficient insulation (covers of Direct FETs are under voltage), at least not for "voltage doubling" by a commutation failure. "High-tech" looks mechanically different.. But it all just on the edge.. ----------------------------------------------- The anti-spark circuit does not behave as it should, that is, there is no continuously extending ramp voltage to the low ESR electrolytic capacitors. Apparently oscillates in between. Very funny, very unusual, very unnecessarily.. The Horse's Mouth Nevertheless, I have lost my bet: Although the voltage rampup when charging the low ESR caps does not run continuously, but that has quite obviously nothing to do with the power-up behavior of the BEC behind it. The fly in the ointment With appropriate load the BEC voltage drops too deep during the pause to be seen above. The result could be that electronics just goes into brownout or other effects. That's it - no reason, no purpose anything to add. Everything is not dramatic. It is solely for the BEC voltage curve at powerup which you may need to keep away from the electrical loads. On the other hand, there are certainly still nothing on the market that could reach the ESCs of K.
__________________
Tom (j-log.eu) Last edited by dl7uae; 12-08-2014 at 06:28 PM.. |
12-08-2014, 02:48 PM | #9 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
We seem to like sports but well..
The anti-spark circuit was really a bad move. It only affects the low ESR caps, low side, as usual. But it is not a "master switch". In addition, it is externally controlled by the MCU. The resolutions of tricky questions are usually simple, here also. Which does not mean that there is a simple solution. Here not as well, at least not within the ESC. I just do it as a picture book. This is the real reason, the rest of the thread could (should?) be deleted actually, except for the Ibec thing in Telme protocol of the JIVE Pro. An interior solution by design changes in KOSMIK or Pro is impossible in my opinion - if, and only then with the "Foxtrot". (that's the way hoaxes are generated ) Fortunately, it is not everyone's problem. "Solution to effect victims" may only be outside of the ESC, see the three options named above. (Unfortunately, our stuff, JLog, CVS, SD, is then also affected.. That was the reason for us to stick our nose in K's marvels.)
__________________
Tom (j-log.eu) Last edited by dl7uae; 12-08-2014 at 05:13 PM.. |
12-08-2014, 06:56 PM | #10 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
I know.. Pictures look less overhead but are not for everyone.
I try myself tomorrow (today, GM ) in a short verbal presentation: What happens? How can it interact with the setup? GN ....... I'm getting a little too lazy. Here is the abstract in German which is still waiting for a translation. Btw: I have the impression that Google Translator has improved significantly since a few days.
__________________
Tom (j-log.eu) Last edited by dl7uae; 12-09-2014 at 04:11 PM.. |
01-07-2015, 12:30 PM | #11 (permalink) | ||
Registered Users
Thread Starter
Join Date: Feb 2008
|
Completion
The story was a lot of support effort or cost post-development time / money (CVS concerning added special protection mechanisms inside). For us it is annoying that in the eyes of the user obviously always first the reputation of our products is in question. Awkward that elucidating the particular cause, although according to the truth, almost amounts to finger pointing to a foreign product. Therefore, we finally HAD TO analyze the actual reason in the inner workings of the ESC - because easier, this happened with a KOSMIK. As a result confirmed / is refined our suggestions for workarounds. The analysis result and mod proposals we presented to K, meanwhile confirmed its CEO the arrival. All right. To simplify our support we make it this way in the future: Quote:
Quote:
With respect to CVS with JLog on KOSMIK or JIVE Pro, we must reserve the right to completely withdraw any support of this setup. It also depends on how well the complex firmware changes to the CVS will be able to ward off future "affliction by voltage madness". --------- 1) KBEC (affects JIVE Pro and KOSMIK) The "BEC power-up problem" can but it must not necessarily happen in a given setup. The cause, which, as just said, acts or does not act ("crazy voltage", see above), is history at latest 100 milliseconds after power-up. After that everything is okay, as regards the BEC. Tentatively, pull out the "+" wire in our "KOSMIK harness" and supply JLog instead via its "JIVE Port" from the BEC voltage of the ESC. If this does not help, there are two alternatives: 1. Remove all load from the BEC before power-up of the ESC, then reconnect the load. 2. Connect a buffer voltage source (battery) with the BEC output (R/C voltage) before power-up of the ESC is done. Thereafter, the buffer can be removed. A third and only clean solution is to use the OptiPower BEC Guard: - Remove "+" wire in "KOSMIK harness" (JLog--KOSMIK/JIVEpro). - Both(!) BEC outputs (master/slave) routed through the BEC Guard. - Supply JLog through its "JIVE Port" from the R/C voltage (BEC voltage) behind the BEC Guard. (In such a case the support of the CVS is not in question.) Information on the "+" wire in "KOSMIK harness" and for powering JLog via the "JIVE port" can be found here at the end of the page. 2) KT The firmware of the JIVE Pro (currently not the KOSMIK) has bugs regarding the protocol with JLog as "TelMe": - Ibec (BEC output current) is nailed to the notional value of 0.9 - Imot (motor current) freezes on the last value when the ESC commutation stops (throttle down). Since not affecting the mAh, it is more cosmetic. Just for the record mentioned: There should have this by now spread that KOSMIK/JIVEpro neither deliver Ubec (BEC voltage) nor the throttle value to "TelMe". This is not a bug, it's a feature. 01/07/2015 (changed Jan 9..11)
__________________
Tom (j-log.eu) Last edited by dl7uae; 01-11-2015 at 03:47 PM.. |
||
01-07-2015, 06:39 PM | #12 (permalink) |
CHC Benefactor
Join Date: Feb 2006
Location: Singapore
|
As always, really appreciate the work you have done helping us use Kontronik products!
|
04-05-2015, 09:30 AM | #13 (permalink) | ||
Registered Users
Thread Starter
Join Date: Feb 2008
|
From Dirk, the guy with whom I examined the Futaba R7003 design flaw.
KOSMIK BEC: Oscillating voltage after power up, - forever if you do not disconnect any load or feed a buffer voltage. Who said that high startup current by certain servos trigger those effects and therefore under rare circumstances? Now Dirk is waiting for arrival of an Opti BEC Guard. ------------------------------ https://www.helifreak.com/showpost.p...45&postcount=1 Quote:
Quote:
Btw - related topic: You cannot use a BEC Guard behind a Kontronik JAZZ (40, 80 etc.). Reason: two times Kontronik @1 BEC's of Kontronik's KOSMIK and JIVE Pro have been a main reason for BG's function "voltage lockout". The BEC voltage curve at power up was to 100% the reason why the turn-on voltage was set to 5 volts. @2 ...and the BEC of a JAZZ provides 4.95V at max
__________________
Tom (j-log.eu) |
||
04-05-2015, 05:19 PM | #14 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
Fortunately this post could be invalidated by my next.
It turned out: This copy of a KOSMIK BEC can not help a BEC Guard. Over a time of about 100 milliseconds after power-up the BEC controller is without control on its feed back input. If the BEC manage it to constantly provide adequate voltage into the subsequent 5V buck regulator and the 5V buck into the subsequent 3.3V buck which supplies the microcontroller (STM32) then the MCU will take control of the BEC through the FB input of its controller - after ~100ms. Before, during the 100ms, the voltage at the FB is unpredictable. Everything can happen.. So also minimum tolerances of resistors can have massive impact by involved non-linear factors on the timeline. The voltage may overshoot beyond the set point but above all, voltage can again drop too deep and that can be a function of load. This is massively so in this particular case. The BEC Guard is not only designed for use on a Kontronik BEC. So its "voltage lockout" waits for the voltage reaches 5V, and after a delay of 2ms it turns the output voltage on with a ramp of another 2 milliseconds. As proved in previous practices sufficient to allow the BEC controller the rest of the waiting time to be decent until his Master (MCU) takes over. Not in this case: This BEC feels already overloaded by 200mA during these 100ms in question, getting extended accordingly. This KOSMIK should be replaced. Not that this would cure something systematically but it's the probable chance to get a "better" copy. Otherwise there are only two options: a) power-up without any load b) to provide a buffer voltage during the time in question, - from a 4..5x NiXX, eventually Eneloop, 2S LiPo or the Ultra Guard with our new firmware It would be possible, but I think the BEC Guard should not be for specific remedies for sick Kontronik BEC's by generally not turning on any load for about 200 milliseconds after power-up.
__________________
Tom (j-log.eu) Last edited by dl7uae; 04-06-2015 at 04:13 AM.. |
04-05-2015, 08:27 PM | #15 (permalink) |
Registered Users
Join Date: Dec 2011
|
I haven't had a chance to try my new BEC guards, (purchased 4 of them), but if I understand your post - the Opti BEC guard will NOT help the BEC problems of the KOSMIK and Jive Pros?
Edit: At least not in all cases. Perhaps I'll just have to wait and see. Sounds as though doing away with the BEC guards and running the Ultra guard with new firmware may be a better solution. Although I do have R2 buffers for all my models, and would prefer to stick with them.
__________________
Tron 7.0 Dnmaic / RAW 420 / Jeti DS-12 Carbon / Futaba 16IZS |
04-06-2015, 03:30 AM | #16 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
Final finally.. Dirk made a mistake: He had a bypass around the BEC Guard.
He forgot to bring JLog BEHIND the BEC Guard and on top he did not cut the red wire in the s.bus2 connection JLog--receiver. So the bypass was: KOSMIK option port "plus" --> JLog --> JLog's voltage regulator 3.3V --> s.bus2 connector "plus pin" (for supply of multiple digital temp sensors or a Unidisplay w/ 3.3V) --> receiver, FBL, R2 buffer(!) etc Now it works. Before: BEC oscillating after power-up. After (with BEC Guard): all ok. So.., indeed this KOSMIK BEC is a particularly bad copy - but the BEC Guard could help here also. The voltage bypass which Dirk just discovered was the problem why it looked incurable.
__________________
Tom (j-log.eu) |
04-06-2015, 03:34 AM | #17 (permalink) | |
Registered Users
|
Quote:
__________________
Henseleit TDR, TDR2, TDF, TDS, TDSF Kontronik, Scorpion ESC, iChargers, VBar, VControl Old hand at Planks Peter |
|
04-06-2015, 03:47 AM | #18 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
See above: It helped, also in this case.
A voltage bypass around the BG was the kidding issue.
__________________
Tom (j-log.eu) |
04-06-2015, 08:24 AM | #19 (permalink) |
Registered Users
Join Date: Dec 2009
|
Tom, i am using BEC guard in 4 of my machines with combination of Kosmik, Jive pro 120 and 80 HV. I am also using Jlog 2.6 with Jeti Tx on all of them. So far so good without any issue initializing or brownouts. For the Jlog, i am wiring kosmik harness and Com port and again everything is working great. Do I have to do the other suggested wiring by cutting + from kosmik harness and routing behind the BEC guard?
__________________
_____________ |
04-06-2015, 09:00 AM | #20 (permalink) |
Registered Users
Thread Starter
Join Date: Feb 2008
|
Well..., better to do that. If you did not experience any problems to JLog (and especially not to the file system on its SD), ok.., but principally better to supply every electronics from the clean voltage out of the BEC Guard.
Just put the removed red wire on COM (JLog's end of the servo cable) on the inner pin ("signal") of JLog's JIVE port - and remote the plus wire from the KOSMIK/JIVEpro harness. Easy to do.
__________________
Tom (j-log.eu) |
Thread Tools | |
Display Modes | |
|
|