HobbyKing.com New Products Flash Sale
Reply
Thread Tools
Old Dec 02, 2010, 04:41 PM
Registered User
France
Joined Nov 2003
278 Posts
ESC response time Graph

Here are the graph.
First set is at a rate of 50 msec, meaning 50 msec between each PWM change.

From Left to right:
BL-CTRL/I2C -- BL-CTRL/PPM -- TPMag8 -- TPMag8 Reflashed
obor is offline Find More Posts by obor
Last edited by obor; Dec 02, 2010 at 04:50 PM.
Reply With Quote
Sign up now
to remove ads between posts
Old Dec 02, 2010, 04:42 PM
Registered User
France
Joined Nov 2003
278 Posts
100 msec between PWM pulse changes.
As you can see, BL-CTRL/I2C and BL-CTRL/PPM are *very* similar. It shows that good response time does not come from the protocol used (I2C or PPM), but the linearity of the controler firmware.
obor is offline Find More Posts by obor
Last edited by obor; Dec 02, 2010 at 04:49 PM.
Reply With Quote
Old Dec 02, 2010, 04:44 PM
Registered User
France
Joined Nov 2003
278 Posts
300 msec
At this rate, the motor has much more time to follow the duty requested, and the RPM vary accordingly.
obor is offline Find More Posts by obor
Last edited by obor; Dec 02, 2010 at 04:52 PM.
Reply With Quote
Old Dec 02, 2010, 05:30 PM
Heli addict
Joined Feb 2007
640 Posts
Thanks Obor for sharing...

btw, can you compare ESC response with "break On" / "break Off"?

Many are using ESC with "break Off"... will "break On" function help on down-spin ?
vladz is offline Find More Posts by vladz
Reply With Quote
Old Dec 03, 2010, 02:48 AM
Registered User
France
Joined Nov 2003
278 Posts
Quote:
Originally Posted by vladz View Post
Thanks Obor for sharing...

btw, can you compare ESC response with "break On" / "break Off"?

Many are using ESC with "break Off"... will "break On" function help on down-spin ?
Yes, thanks for the suggestion. Let me add it to a todo list:

- break ON
- governor ON (why not ?) Governor is probably a bad idea.

Anything else that would be interesting ?
obor is offline Find More Posts by obor
Last edited by obor; Dec 04, 2010 at 02:44 AM.
Reply With Quote
Old Dec 03, 2010, 07:29 PM
Tri-Quad-Hexa-Octo-copters!!
United States, TX, San Antonio
Joined Feb 2007
14,417 Posts
Hi Obor,
Perhaps motor timing in the typical High, Medium and Low settings using 750kv and 1200Kv motors?
Cheers,
Jim
jesolins is offline Find More Posts by jesolins
Reply With Quote
Old Dec 04, 2010, 02:49 AM
Registered User
France
Joined Nov 2003
278 Posts
Thanks Jesolins,
So the list is now:
- break ON,
- Motor Timing
(Governor removed)
obor is offline Find More Posts by obor
Reply With Quote
Old Dec 06, 2010, 04:50 PM
Registered User
France
Joined Nov 2003
278 Posts
New test tonight with a different PPM signal. Not yet tested break on and timing, I'll wait to receive new tg plush I ordered before doing the test.
obor is offline Find More Posts by obor
Reply With Quote
Old Dec 07, 2010, 11:17 AM
Registered User
Italy
Joined Dec 2010
4 Posts
Hi guys,

my name is Marco and i'm italian (so forgive my bad english).
I was wandering on internet searching some thread about esc's response time when i found this great discussion. I wanna say thank you to kapteinkuk and obor for publishing all this data. I'm interested in modelize a quadcopter on Matlab and this data, expecially obor's data, are wery helpful to describe latency of ESC firmware+motor inertia.

Since i read you are in "suggestion time" i wanna propose to represent on graph RPM behaviour after a single PWM step (like kapteinkuk graph), so graphs should result more easy to read and understand wich is the fastest ESC firmware. Can be considered for example 2 single PWM step: a big one (10%-100%) and a small one (50%-70%).
Quote:
Originally Posted by obor
Now what controls should be applied, to best represent a Kopter behavior:

- 10%-100%
- 40%-60%
- 50%-70%
?
pando14 is offline Find More Posts by pando14
Reply With Quote
Old Dec 07, 2010, 12:52 PM
Registered User
France
Joined Nov 2003
278 Posts
Hi Marco,
Welcome to this thread

Quote:
Originally Posted by pando14 View Post
...
Since i read you are in "suggestion time" i wanna propose to represent on graph RPM behaviour after a single PWM step (like kapteinkuk graph), so graphs should result more easy to read and understand wich is the fastest ESC firmware. Can be considered for example 2 single PWM step: a big one (10%-100%) and a small one (50%-70%).
Yes, thanks this is a very good suggestion.
But doing juste one single step up or down does not show all. For instance, how the firmware handle PWM changes during the rampup or rampdown ?

So may be, to make it more clear, graphs could be simplified that way:
- one shot 10 - 100%
- one shot 10 - 100 - 10% with 100msec between pulse
- one shot 50 - 70%
- one shot 50 - 70 - 50% with 100msec between pulse
?
obor is offline Find More Posts by obor
Reply With Quote
Old Dec 09, 2010, 11:59 AM
Registered User
Italy
Joined Dec 2010
4 Posts
[QUOTE=obor]
pando14 is offline Find More Posts by pando14
Last edited by pando14; Dec 09, 2010 at 12:08 PM. Reason: error while writing
Reply With Quote
Old Dec 09, 2010, 12:05 PM
Registered User
Italy
Joined Dec 2010
4 Posts
You right..I completely forget rampdown (stepdown) response
With "rampdown"(up) do you mean stepdown(up)? however rampdown(up) is simply a lot of step one after another so maybe you can simplify your test doing:
-just one shot 10 - 100% ---> wait RPM adjustment (100ms could be too little) ---> just one shot 100 - 10%

-just one shot 50 - 70% ---> wait RPM adjustment (100ms could be too little) ---> just one shot 70 - 50%

I wrote you this things because the final objective is to find the best ESC response (isn't it?) and in your previous graphs i can't understand it clearly because i can't understand time scale (i think because there are too much pulses so time scale become smaller or maybe because I can't read it). So maybe a simply step (up and down) response can show it better.

Quote:
Originally Posted by obor
Currently, my settings for RPM measurements are the following:
- 6 magnets on the motor
- measurements are done every 25msec
One measure every 25ms (40Hz) seems to be not enough because time responses in kapteinkuk's graph are:
1) 137ms with big pulse and 50Hz pwm
2) <25ms with big pulse and 400Hz pwm

With 1measure every 25ms we can capture 5-6 points on rising (falling) edge of the RPM behaviour in first case and no one in the second case. With smaller pulse (like 50-70% ), the rising edge (falling) time response will be even more little!

Maybe the motor inertia could help us and delay time response but i'm a little bit doubtful..what do you think about this?
pando14 is offline Find More Posts by pando14
Reply With Quote
Old Dec 09, 2010, 03:06 PM
Registered User
France
Joined Nov 2003
278 Posts
Quote:
Originally Posted by pando14 View Post
You right..I completely forget rampdown (stepdown) response
With "rampdown"(up) do you mean stepdown(up)? however rampdown(up) is simply a lot of step one after another so maybe you can simplify your test doing:
-just one shot 10 - 100% ---> wait RPM adjustment (100ms could be too little) ---> just one shot 100 - 10%

-just one shot 50 - 70% ---> wait RPM adjustment (100ms could be too little) ---> just one shot 70 - 50%

Quote:
I wrote you this things because the final objective is to find the best ESC response (isn't it?) and in your previous graphs i can't understand it clearly because i can't understand time scale (i think because there are too much pulses so time scale become smaller or maybe because I can't read it). So maybe a simply step (up and down) response can show it better.
In previous graph, the X axes was time. Between each PWM change there is 50, 100 or 300 msec depending on the graph. So you don't need much to bother with the time, it is only interesting to know that time between PWM pulse is constant.

[QOUTE]
One measure every 25ms (40Hz) seems to be not enough because time responses in kapteinkuk's graph are:
1) 137ms with big pulse and 50Hz pwm
2) <25ms with big pulse and 400Hz pwm
To clarify what is measured, every 25msec the number of hits (magnet in front of the sensor) is returned. Below 25msec, the number of hits become fairly small, and approximation increase. There are 6 magnets on my motor, the accuracy is 1/6 turn. It may worth trying below 25msec. I looks there is no direct mapping between current consumption and RPM...
Quote:
With 1measure every 25ms we can capture 5-6 points on rising (falling) edge of the RPM behaviour in first case and no one in the second case. With smaller pulse (like 50-70% ), the rising edge (falling) time response will be even more little!
At highst turn rate, the motor speed approximately 10000 RPM (to simplify). It means 166 rotation per second, so 0.166 per msec. 25 msec is 4.16 turn.
Ok, may be it could be reduced down to 2turns at full speed, which is 1 turn at half speed, so 12msec would be worth trying.

Quote:
Maybe the motor inertia could help us and delay time response but i'm a little bit doubtful..what do you think about this?
Some people have emit doubts about the fact that no propeller was used during that test. So, "with propeller" is to add to the test list.
obor is offline Find More Posts by obor
Reply With Quote
Old Dec 14, 2010, 05:48 PM
Registered User
Italy
Joined Dec 2010
4 Posts
I wanna underline 3 point:
1)
Quote:
Originally Posted by obor
In previous graph, the X axes was time. Between each PWM change there is 50, 100 or 300 msec depending on the graph. So you don't need much to bother with the time, it is only interesting to know that time between PWM pulse is constant
I had understood time was on X axis but your graphs can't give an objective method to understand what's the best ESC time response (like instead kapteinkuk's graphs) so my suggestion is to calculate time response after 1 step input (like kapteinkuk's graphs). You can also calculate time response after 1 ramp input but show on a graph a step response is more clear.

2)
Quote:
Originally Posted by obor
But doing juste one single step up or down does not show all. For instance, how the firmware handle PWM changes during the rampup or rampdown?
Calculate time response after a stepup (stepdown) input give you the same info than response after rampup (rumpdown) because in digital mode a ramp is made by steps! So a fast time response after a stepup (stempdown) means a fast time response after a rampup (rumpdown)

3)
Quote:
Originally Posted by obor
to clarify what is measured, every 25msec the number of hits (magnet in front of the sensor) is returned. Below 25msec, the number of hits become fairly small, and approximation increase. There are 6 magnets on my motor, the accuracy is 1/6 turn. It may worth trying below 25msec.
Ok, I understood why you can't reduce time between measures, i had the same problem in another project! What i mean with
Quote:
One measure every 25ms (40Hz) seems to be not enough
is that for example in your graph of 50msec between PWM change, with your measure frequency, you maybe can't correctly describe RPM behaviour. I divided you graph to show better the period and in my example (i zoom - sorry for very very bad quality - period 3) the blue line indicate the real RPM behaviour and red cross data captured with you measures(ok, this is a very unfortunate case but it's just to explain). So your measure show no RMP change but in reality there's change. In addition to this problem there's the RPM approximation caused by the low accuracy (1/6 turn). The sum of this two problem is clear in your graph because the RPM behaviour is different in every period (for example from 1 to 9) instead it should be the same because in every period you have the same input sequence.
At the end, with this measure frequency you couldn't appreciate, at "high" frequency (for example 50ms between pwm changes), the difference between two esc response.
I wanna say that I don't want to destroy your work, instead i want to help you to find a method to compare different esc response .. maybe we have to think about using current measure.

P.S. Does anyone know something about turnigy AE - 20 esc? In the description is wrote "The new CPU provides crisper throttle response: smooth, linear, quick and precise throttle response."..it seems the response to our problems
pando14 is offline Find More Posts by pando14
Last edited by pando14; Dec 14, 2010 at 06:05 PM. Reason: error correction
Reply With Quote
Old Dec 15, 2010, 12:42 PM
Registered User
France
Joined Nov 2003
278 Posts
Quote:
Originally Posted by pando14 View Post
Calculate time response after a stepup (stepdown) input give you the same info than response after rampup (rumpdown) because in digital mode a ramp is made by steps! So a fast time response after a stepup (stempdown) means a fast time response after a rampup (rumpdown)
I agree with you if the rampup(or rampdown) is reached before the next PWM duty. If not, then the way firmware handle rampup or down may interfere.

Quote:
Ok, I understood why you can't reduce time between measures, i had the same problem in another project! What i mean with
is that for example in your graph of 50msec between PWM change, with your measure frequency, you maybe can't correctly describe RPM behaviour.
Yes, you are right, with small PWM period changes like 50ms, the RPM capture every 25msec is not accurate enough.

Quote:
I divided you graph to show better the period and in my example (i zoom - sorry for very very bad quality - period 3) the blue line indicate the real RPM behaviour and red cross data captured with you measures(ok, this is a very unfortunate case but it's just to explain). So your measure show no RMP change but in reality there's change. In addition to this problem there's the RPM approximation caused by the low accuracy (1/6 turn). The sum of this two problem is clear in your graph because the RPM behaviour is different in every period (for example from 1 to 9) instead it should be the same because in every period you have the same input sequence.

At the end, with this measure frequency you couldn't appreciate, at "high" frequency (for example 50ms between pwm changes), the difference between two esc response.
I wanna say that I don't want to destroy your work, instead i want to help you to find a method to compare different esc response .. maybe we have to think about using current measure.
I agree with this, with small period like 50msec ,measurement are not accurate enough. But do we need to measure that low ? I did 100 msec measurement, but could try also 150 or 200 msec between PWM changes.

If we consider that it takes approx 4msec for a turn a full speed, then with 6 magnets it gives:
- in 4 msec, Magnet seen vary between 0 - 6
- in 8 msec, it gives 0-12
For RPM varying between 2000 and 3000, that means 0-3 to 0-4/5 in 4 msec, which 1 or 2 magnet more or less per turn, which is not enough.
I'll try to double the number of magnets.
obor is offline Find More Posts by obor
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Discussion Testing ESC's? pdown1231 Power Systems 3 Feb 23, 2010 07:02 PM
Help! Hobbyzone Super Cub LP No Response from ESC/RX bladerunner2 Power Systems 0 Nov 14, 2009 12:30 AM
Help! Weird motor/ESC response to differant batteries. c/f Power Systems 19 Oct 26, 2009 06:49 PM
Video video of cyclic response testing : EVO450 GIGA_RC Electric Heli Talk 3 Sep 02, 2008 12:18 PM
Question ESC Response at low throttle. RGinCanada Dock Talk 14 Jan 07, 2005 01:36 PM