RunRyder - TX/RX eCCPM Latency Test Results

rrTV-PHOTO

New HD TV

HOME rrTV-PHOTO GALLERIES MY GALLERY HELP-FAQ

myHOME PM pmRR MEMBERS 451 ONLINE 20 EVENTS SEARCH REGISTER START HERE

54 pages [ << < 37 38 ( 39 ) 40 41 NEXT >> ]

85436 views

POST REPLY

Mikado Modellhubschrauber	.	GrandRC	.	CanoMod
	.		.

Radio - Servo - Gyro - Gov - Batt > TX/RX eCCPM Latency Test Results

Four Stroker
Veteran
Location: Atlanta

Well, JKos could step the servo input through say 4096 steps and measure the servo output (voltage at the pot) and do a quick analysis on his logic analyzer.

EDIT: Logic analyzer is the wrong tool here. Need some kind of A/D and save the data.

02-23-2008 07:58 PM

Four Stroker
Veteran
Location: Atlanta

AirwolfRC, The answer might be in the gear train slop. If you take a Hitec TG servo and program it to the least deadband, then it will howl at 300 Hz or so as the motor bounces back and forth between two (or more) motor poles. The slop in the gear train means that the actual servo output will not actually be following. It is possible that the servo output, connected to the pot, could settle out and find a more accurate position - a stochastic process kind of thing.

Experiment:

1) Sweep the servo input through 4096 positions or more.

2) Synchronize and sample the servo pot voltage and save as a data set. Take > 10 samples at 14 bits or so for each input step.

3) Average the voltage at each step.

Look and see what happens. Servo should be driven by a stable power supply.

If you adjust the deadband, like Kraft did, so that the servo does not buzz, then the servo step counting thing works pretty well. Proline used to set the deadband so that the servos buzzed continuously in an attempt to increase actual resolution. With Hitec you have a choice.

02-23-2008 08:34 PM

JKos
Elite Veteran
Location: City of California in the state of Maryland

How certain are we of the gear ratios? Obviously it varies between servos, but has anyone actually counted it out on a specific model or two?

Under load, the motor is "continuously" driven; so who's to say that it has to "stop" at pole positions.

- John

MSH Protos

02-23-2008 08:49 PM

hootowl
Elite Veteran
Location: Garnet Valley, Pa.

Since the motor has to turn many many revolutions to turn the output shaft just a little, don't you have to multiply that 200:1 3-pole x the number of revs it takes to turn the output shaft one rev to get your resolution. In otherwords if it takes say 300 revolutions of the motor to turn the output shaft one rev, don't you multiply the pole motor resolution times the number of revs the motor has to make to get your available resolution?

Member Bog Troll Club #1

02-23-2008 09:29 PM

HOMEPAGE

JKos
Elite Veteran
Location: City of California in the state of Maryland

I opened up a Hitec HS-81MG servo. It has a dead band of 8 us and is listed as having a 3-pole motor. First bit of data is that the motor will cog at 12 positions. I have confirmed that when driving the servo, the motor does indeed cog one position at a time (with no load).

This is a good motor for the studio as it has poor specs, the gear train stays altogether, and the motor pinion is fully exposed with the case apart.

Now on to the gear ratio. Fun, fun...

- John

MSH Protos

02-23-2008 09:41 PM

AirWolfRC
rrProfessor
Location: 42½ N, 83½ W

Hmmm, dead band of 8us ? 1ms / 8us = 125 increment resolution

I know that is not a good characterization but an indicator either way.

02-23-2008 09:48 PM

HOMEPAGE

JKos
Elite Veteran
Location: City of California in the state of Maryland

> Hmmm, dead band of 8us ? 1ms / 8us = 125 increment resolution

Here is what I observed... Via the dead band test on the Vexa ServoXciter EF, the setting which causes the servo to start buzzing indeed matches the spec for the HS-81MG. However, when simply driving it with a adjustable pulse width signal, it moves well before a 8 us change in pulse width. Just under 2 us. So, you could say it is somewhere around 512 step resolution by this measurement.

So, that tells me that the dead band is not directly related to resolution. At least not on the this servo.

Here is another aspect to consider... There are two loops going on, right? One is the servo's response to the input pulse width. The other is the servo's response to a change in the pots position. Would this not create the possibility of three categories?

1) Servo allows output position to deviate more than its input pulse width resolution.
2) Servo output position deviation allowance matches input pulse width resolution.
3) Servo holds output position tighter than its input pulse width resolution.

- John

MSH Protos

02-24-2008 01:43 AM

Four Stroker
Veteran
Location: Atlanta

The output pot goes directly to a pulse generator. When the internal pulse width and the servo input command pulse width differ by more than the deadband, the motor is activated. When driving the servo in one direction, it will indeed move in much smaller increments. For example, the servo stops when the pulse difference is 7 us. If you change the input command by 2 us, the difference is now 9 us ( > 8 us )and the motor moves. If you back up, the dead band will give you a hysteresis. The deadband has to do with absolute positioning accuracy and not really step size in one direction.

The servo pot is also very noisy, so the servo will always buzz if the deadband is too small.

02-24-2008 10:54 AM

AirWolfRC
rrProfessor
Location: 42½ N, 83½ W

Four Stroker, yep, exactly.

One thing still sure is that servo will see little improvement in resolution if any from a 2048 resolution radio.

One way to test servo resolution is to setup a test where the servo is getting a slowly increasing pulse width and the servo has a preload on the output shaft against the servo travel and then measure the output arm movement.

This would eliminate dead band and backlash.

02-24-2008 05:09 PM

HOMEPAGE

tadawson
Elite Veteran
Location: Lewisville, TX

I did a calculation on a thread somewhere (can't remember where . . . ) considering the distance of motion on a 1" arm on 2048 for a typical servo, and it was something like 5/10000 of an inch (yes, TEN thousand) and I doubt that most folks linkages have less slop than that, thus further negating any benefit of 2048 . . . the fanboys say "but you can feel it", but I have to wonder if what they feel is really the burn in their wallet from the upgrade, more than much actual difference . . .

- Tim

02-24-2008 06:42 PM

GimbalFan
Elite Veteran
Location: Copter County, Nv

You could also consider the comparison of tx stick movements. My Stylus' sticks move about 1.8". Divide that by 1024 = 0.0018" = 1/570" = 1/22mm of movement per step of theoretical system resolution. That's an incredibly small amount of stick movement in which to claim to be able to feel the difference.

2048 steps? Fuggedaboudit. That can be nothing more than buzzword hype.

Quite frankly I believe 512 or even 256 would be plenty tight enough for all but the most precise types of flying, or for all but the very best competitive pilots.

op-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-thwõp-t

02-24-2008 06:52 PM

tadawson
Elite Veteran
Location: Lewisville, TX

That was pretty much the conclusion I came to, but I got ridiculed by the fanboys who didn't want to be bothered with facts . . . .

- Tim

02-24-2008 06:57 PM

AirWolfRC
rrProfessor
Location: 42½ N, 83½ W

I have no doubt that a high end (2048 resolution) radio can and will show better control by giving a "more connected feeling".

But you won't see that unless you also use high end servos, even though the output push rod is only moving 1/2 thousandth of an inch per increment of resolution. (along with minimal latency)

02-24-2008 07:28 PM

HOMEPAGE

hootowl
Elite Veteran
Location: Garnet Valley, Pa.

Quote
There is something wrong with the 200:1, 3-pole motor concept as even the cheapest servos I have are clearly much better than 150, 300, or even 600 steps. - John

How many steps does is motor itself capable of in one motor revolution?

Member Bog Troll Club #1

02-24-2008 09:16 PM

HOMEPAGE

Four Stroker
Veteran
Location: Atlanta

Well let's back up and start over a little bit. A DC motor will move in infinistesimal steps if controlled properly. For example, a nearly stalled motor will slowly rotate and if you remove the drive voltage at the right time, it will obviously stop there. On the other hand, everyone has seen a motor cog when started up slowly. There is a non-linear jump when the rotor commutates - changes which plates the brushes are on. Modelling motors on a small scale is very complicated.

So there is really no answer to your question.

What we are discussing here is the total servo and how it is set up. In the not too distant past, 9202 vintage, servos were purely analog and had a linear analog feedback loop. They moved in little steps. Hold one up to your ear and listen as you move the throttle stick as slowly as you can. Do the same thing with a new high end coreless servo with a 2048 system and the difference is obvious. There are a lot more smaller steps. This is a stream of consciousness (aka BS) thread about why that is.

With digital servo amplifiers, performance can improve dramatically but could we tell the difference ?

02-24-2008 09:50 PM

JKos
Elite Veteran
Location: City of California in the state of Maryland

No one can tell me that I have not experienced smoother servo motion with a 2048 radio. No one can tell me that because I have seen it with my own eyes and felt it in the servo motion. There was even a difference in the sound as you slowly moved a stick. An example of a servo on which it is quite obvious is the JR DS3421. But, anyways...

There is now a system readily available which will allow one to directly experience the difference between 2048 and 1024. Just take an X9303 and setup a heli using an AR9000 series receiver (or R921), fly it, then swap in an AR7000. No change in latency, no change in CCPM characteristics, no change in heli setup. The only change is resolution.

> but I have to wonder if what they feel is really the burn in their
> wallet from the upgrade

Last I checked, upgrading to 2048 (or happening to get 2048 as bonus with a new radio) is not expensive these days.

> How many steps does is motor itself capable of in one motor
> revolution?

The 3-pole motor in this HS-81MG has 12 cog positions and I can see it going one at time on occasion.

- John

MSH Protos

02-24-2008 10:15 PM

hootowl
Elite Veteran
Location: Garnet Valley, Pa.

A three pole DC motor can have 6 or more "pulses" depending on the magnet configuration.

Look at the interactive diagram and count the number of north/north-south/south occurrences in one revolution.

Note it does 12 steps in the interactive diagram for this motor.

http://www.solarbotics.net/starting...1_dcmotor2.html

12 x 200 = 2400 steps per rev on the output shaft.

360/2400
.15 degrees resolution capable.

Member Bog Troll Club #1

02-24-2008 10:31 PM

HOMEPAGE

AirWolfRC
rrProfessor
Location: 42½ N, 83½ W

A brushed motor with an odd number of poles really has NO cog positions.
The reason for three poles is to eliminate "cog" positions.
When voltage is applied to the armature, it will try to turn, no matter what position it's in.

02-24-2008 10:32 PM

HOMEPAGE

hootowl
Elite Veteran
Location: Garnet Valley, Pa.

Yes but it does have positions where there are pulses.

Member Bog Troll Club #1

02-24-2008 10:38 PM

HOMEPAGE

AirWolfRC
rrProfessor
Location: 42½ N, 83½ W

Not really.
If a pole happens to be lined up with a field magnet, the other two poles will be straddeling the other field magnet and one of those poles will be pulling to that magnet and the other will be pushing.

The "coging" you may feel when rotating a non-energized motor is the magnets attracting the iron in the poles. A coreless morot, withour iron in the core, will have no coging.

02-24-2008 10:41 PM

HOMEPAGE

54 pages [ << < 37 38 ( 39 ) 40 41 NEXT >> ]

85436 views

POST REPLY

Futaba-RC	.	A Main Hobbies	.	Boca Bearings
	.		.

Radio - Servo - Gyro - Gov - Batt > TX/RX eCCPM Latency Test Results

PRINT TOPIC

Advertisers

Subscribe to This Topic