RunRyder RC
WATCH
 3 pages [ <<    <     1     ( 2 )     3     NEXT    >> ] 3025 views POST REPLY
HomeRC & PowerAircraftHelicopterFlybarless Rotor Head SystemsOther › Flybarless Controller Latency Testing
02-19-2013 04:25 PM  5 years agoPost 21
JKos

rrProfessor

Redondo Beach, CA

My Posts: All  Forum  Topic

I will agree that there is certainly more going on in the control loop when testing these FBL controllers. Changes to the control loop can have more impact on feel then any change in latency. However, if the controller takes time to transfer your direct inputs, such as collective, to the servos, then that's the same latency that I was measuring with just transmitters and receivers.

MrMel,
For sample, VBar 4.0 vs 5.0, same speed on control loop, same hardware, one frame every 7ms yet the 4.0 feels like spagetthi compared to 5.0 when flown.
Do you know with 100% certainty that there was no change to the latency of direct inputs? Operating at 7 ms frames does not imply low latency as shown by the test results above with 5 ms frames out of the AR7200BX. Even with 5 ms frames, it still took up to 53 ms before any change started to happen.

AWittleWabbit,
Was pitch boost off?
I'll test with it on as well. It may be fun to plot the collective response in each setting. I'll also test cyclic inputs.

- John

PM  EMAIL  GALLERY  Attn:RR  Quote
02-19-2013 06:28 PM  5 years agoPost 22
MrMel

rrProfessor

Gotland

My Posts: All  Forum  Topic

Do you know with 100% certainty that there was no change to the latency of direct inputs? Operating at 7 ms frames does not imply low latency as shown by the test results above with 5 ms frames out of the AR7200BX. Even with 5 ms frames, it still took up to 53 ms before any change started to happen.
My point is, it does not matter so why bother.

Like you see above, 53 millisecond, it actually take another 100 before the heli turns at least (physical movement) so latency has no point in FBL solutions if the Gyro control loop is not there to actually realize the movement anyway.

Now, even if the next FBL system you test show 80ms for "same test", it can be WAY faster in reality (actually feel hand to heli movement)

Or a comparison to measure latency of a power steering (progressive) on a car without knowing the speed it drives.
10ms or 100ms might not differ at all if the first car is driving 10mph and second 100mph

Gone fishing..or hunting..or something
My site: http://heli.dacsa.net - VBar videos and more

PM  EMAIL  GALLERY  Attn:RR  Quote
02-19-2013 09:42 PM  5 years agoPost 23
Four Stroker

rrElite Veteran

Atlanta

My Posts: All  Forum  Topic

MrMel, why bother reading the thread if you don't have any interest in the results ? Your fixation on trashing this thread indicates to me that you have a vested interest in a system that you are afraid will test poorly. In the USA we call this a TEAM pilot.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-19-2013 09:53 PM  5 years agoPost 24
Justin Stuart (RIP)

rrMaster

Plano, Texas

My Posts: All  Forum  Topic

John, can you please post some pictures and a description of exactly how you are doing your testing?

Avant RC
Scorpion Power Systems
Thunder Power RC
Kontronik Drives

PM  EMAIL  GALLERY  Attn:RR  Quote
02-19-2013 09:57 PM  5 years agoPost 25
MrMel

rrProfessor

Gotland

My Posts: All  Forum  Topic

MrMel, why bother reading the thread if you don't have any interest in the results ? Your fixation on trashing this thread indicates to me that you have a vested interest in a system that you are afraid will test poorly. In the USA we call this a TEAM pilot.
Has nothing to do with that but I had this discussion with manufacturers when it was talk about center speed (invention of paddle simulation on vbar), center speed of CGY750 being fast etc, without doing anything about latency.

Hence I know from that discussion that my comparison to powersteering is quite correct.

But if you want numbers of "something", go ahead.

Gone fishing..or hunting..or something
My site: http://heli.dacsa.net - VBar videos and more

PM  EMAIL  GALLERY  Attn:RR  Quote
02-19-2013 10:21 PM  5 years agoPost 26
rexxigpilot

rrProfessor

Florida

My Posts: All  Forum  Topic

I see MrMel's point. For those among us who deal in numbers, calculations and computations on a daily basis, we can put the raw data into perspective. Those who don't will run with a raw number to justify that system A is better than system B, when the reality is there are many much more important factors in a good FBL system than simple latency. This issue of being scientifically or engineering minded or not was clearly a problem with John's original TX Latency thread. Just the fact that most on RC heli forums (I'd say around 90%) don't even understand what a PID control loop is, or know the difference between feed-back versus feed-forward is perhaps reason enough for Fredrik's concern.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-19-2013 11:41 PM  5 years agoPost 27
BobOD

rrElite Veteran

New York- USA

My Posts: All  Forum  Topic

Yeah but is collective part of the control loop?
I haven't decided if I think these differences are important or not but I do find it somewhat interesting to see the information, discuss, and then decide.
Food for thought.
Lets say you decide to make a move. You have (just guessing #s here);
Delay before your finger moves - 20ms
TX/RX delay - 50ms
Mechanical delay - 100ms (seems high to me but go with it for now)
Delay before your brain registers what your eye saw - 20 ms

Total delay 190ms

If the tx/rx was 100ms, then the total is 240ms. That would be a 26% increase in delay. That might be noticeable.
However, if the tx/rx was 60ms, then the total is 200ms. That would be a 5% increase in delay. Not so noticeable.

Team POP Secret

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 12:47 AM  5 years agoPost 28
AWittleWabbit

rrElite Veteran

O.C., CA

My Posts: All  Forum  Topic

From a latency standpoint, collective is probably the best to measure. Aside from pitch boost, I can't think of anything the control loop would do with the signal.

Heli-itis sufferer.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 02:42 AM  5 years agoPost 29
Four Stroker

rrElite Veteran

Atlanta

My Posts: All  Forum  Topic

I did not see any problems with John's other thread and I do not see any problems with this thread. It's a measurement. The only possible reason to object to a measurement is that it might conflict with your agenda - whatever that might be. You can comment on what you think the measurement means or does not mean all that you want. Declaring that 90% of RC Heli Pilots are not smart like you is really too much.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 03:01 AM  5 years agoPost 30
BobOD

rrElite Veteran

New York- USA

My Posts: All  Forum  Topic

Rex didn't say "not smart" so it shouldn't be taken that way. Some don't know and don't even want to know. It doesn't need to be taken as insult. It seems a valid opinion to me.

Team POP Secret

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 03:20 AM  5 years agoPost 31
AWittleWabbit

rrElite Veteran

O.C., CA

My Posts: All  Forum  Topic

I'll test with it on as well. It may be fun to plot the collective response in each setting. I'll also test cyclic inputs.
That would be great John. I'm curious whether is akin to neg expo(JR) or if it's feedfoward.

Heli-itis sufferer.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 04:54 AM  5 years agoPost 32
JKos

rrProfessor

Redondo Beach, CA

My Posts: All  Forum  Topic

Here is the difference between the collective response with Pitch Boost turned Off and Very High. No other changes were made.

DISCLAIMER: This is a single test case and does not represent the response of the AR7200BX Pitch Boost function under all conditions.

- John

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 05:04 AM  5 years agoPost 33
AWittleWabbit

rrElite Veteran

O.C., CA

My Posts: All  Forum  Topic

So the ramp rate is the same, just some overshoot baked in. Interesting . It's odd that the pulse width came back down to 1.6ms. I wonder if it would behave differently if your step command was less. Thoughts?

Heli-itis sufferer.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 01:20 PM  5 years agoPost 34
Four Stroker

rrElite Veteran

Atlanta

My Posts: All  Forum  Topic

Now is the time for smart guy comments on this implementation of Boost as MEASURED by John. What were you expecting the graph to look like ? What would the graph look like if you implemented Boost ? What problem does Boost mitigate in the first place.

Discussion should now begin hopefully not interupted too much my TEAM marketing.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 03:54 PM  5 years agoPost 35
Justin Stuart (RIP)

rrMaster

Plano, Texas

My Posts: All  Forum  Topic

I appreciate the graph of the AR7200BX Pitch Boost, but exactly what are we looking at here?

Servo latency is pretty easy to understand. You're measuring the time between when I wiggle the stick on my TX and the time the servo responds. But with a FBL system it is more complex.

Please help interpret that graph for me.

Avant RC
Scorpion Power Systems
Thunder Power RC
Kontronik Drives

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 04:03 PM  5 years agoPost 36
rexxigpilot

rrProfessor

Florida

My Posts: All  Forum  Topic

That graph tells a whole bunch if you understand what you are looking at. What surprised me is that the "Boost" doesn't boost at all. It actually attenuates. This might help maintain headspeed by reaching commanded (directly proportional to TX command) collective then backing down on pitch before the blades bog. I guess the "boost" is an illusion of increased collective pitch created by conserving the headspeed.

The latency was unaffected. Also, the implementation of it is feed-forward. In other words, preprogrammed to go to full commanded pitch then reduce pitch by a set amount.

My implementation would overshoot the commanded collective pitch by a preprogrammed amount, then reduce pitch back to commanded.

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 04:34 PM  5 years agoPost 37
JKos

rrProfessor

Redondo Beach, CA

My Posts: All  Forum  Topic

Gentlemen,
Before we make too many conclusions based on the single graph above of a single test case, please allow me to do some further testing. For example, I need to test small versus large collective inputs. How does it vary with the speed of the input? How does it respond to back to back quick inputs?

The instructions clearly indicate there are limits and interactions to how Pitch Boost works/responds. I can only take measurements on the bench under certain conditions.
What surprised me is that the "Boost" doesn't boost at all. It actually attenuates.
Does collective pitch range need to be adjusted when turning on Pitch Boost? I don't know yet. I ran out of time last night to really check into such things. Clearly the unit should not exceed certain throw settings due to physical travel limits.

Clearly, any of this testing has many caveats and should not be construed as being as straightforward as transmitter/receiver latency.

Thanks,
John

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 06:28 PM  5 years agoPost 38
JKos

rrProfessor

Redondo Beach, CA

My Posts: All  Forum  Topic

John, can you please post some pictures and a description of exactly how you are doing your testing?
Justin,
The testing procedure is the same as I perform for transmitter/receiver latency testing. In short, I inject an input into the radio such that it thinks a stick has been moved. I monitor the device under test's outputs (pulse widths) to detect a change in the pulse width. The amount of time between the injected input and the change in pulse width is the latency.

In longer form, the input injection is accomplished via a microcontroller which generates the input changes as a digital output. That output is injected into the radio axis potentiometer via an injection resistor which is valued appropriately for the radio under test. To the radio, the stick is being moved between two positions.

The microcontroller can either freewheel and generate a square wave at about 3 Hz or be synched to a trigger input. In synched mode, after a preset hold time, the microcontroller will delay an adjustable amount of time after the trigger input before changing the state of the output. An output of the device under test is used as the trigger input. When testing synchronous systems (i.e., a system for which a sample of the stick position corresponds to one output and that time is constant), using the synched mode allows for honing in on the minimum and maximum latency by varying the delay.

For asynchronous systems (i.e., some radios and most likely all of the flybarless systems), the timing of the radio and output of the device under test slide relative to each other. Therefore, one stick position measurement does not correspond to any one output and the stick sample to output latency is not constant. In this case, the freewheel mode is used.

Data is collected on a computer via a logic analyzer and interpreted via MATLAB. As many data points can be collected as desired for statistical analysis and to insure the min and max latency have been hit. One-thousand samples tends to provide good results.

- John

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 07:21 PM  5 years agoPost 39
Justin Stuart (RIP)

rrMaster

Plano, Texas

My Posts: All  Forum  Topic

Thanks!!

Avant RC
Scorpion Power Systems
Thunder Power RC
Kontronik Drives

PM  EMAIL  GALLERY  Attn:RR  Quote
02-20-2013 07:33 PM  5 years agoPost 40
Four Stroker

rrElite Veteran

Atlanta

My Posts: All  Forum  Topic

The IEEE had an article comparing 19 products some years ago. The difference between MATLAB and the others - marketing.

PM  EMAIL  GALLERY  Attn:RR  Quote
WATCH
 3 pages [ <<    <     1     ( 2 )     3     NEXT    >> ] 3025 views POST REPLY
HomeRC & PowerAircraftHelicopterFlybarless Rotor Head SystemsOther › Flybarless Controller Latency Testing
 Print TOPIC  Make Suggestion 

 13  Topic Subscribe

Monday, December 17 - 9:57 am - Copyright © 2000-2018 RunRyder   EMAILEnable Cookies

The RC discussion world needs to consolidate. RR is now one choice for that. Its software is cutting edge. It hosts on-topic advertising. Help RR increase traffic buy making suggestions, posting in RR's new areas (sites) and by spreading the word.

The RunRyder Difference

• Category system to allow Rep/Vendor postings.
• Classifieds with sold (hidden) category.
• Classifieds with separate view new.
• Answer PMs offsite via email reply.
• Member gallery photos with advanced scripting.
• Gallery photo viewer integrated into postings.
• Highly refined search with advanced back end.
• Hosts its own high end fast response servers.
• Hosts thousands of HD event coverage videos.
• Rewrote entire code base with latest technology.
• No off-topic (annoying) click bait advertising.
Login Here
 New Subscriptions 
 Buddies Online