AlexanderMikhailov

Hitec HS-5125 digital servo. Datasheet states 1us dead-band width. Very nice number as it gives user 1400 steps and even exceeds the possibilities of 1024PCM. Actually only 2048PCM could fully exploit 1us resolution advertised for HS-5125.

I personally like these servos and use them a lot. However my feeling was that HS-5125 actual dead-band was wider than advertised 1us. Beeing curiuos I decided to test real servo resolution. The result was discouraging.

Test bed.
Hitec HS-5125 v.1.04 digital servo
Hitec HFP-10 servo tester/ digital servo programmer

Results.

1. DB width test. Servo comes to user with factory default dead-band setting DB=2 (=6us). When I asked MikeMayberry from Hitec about my finding he explained it was a "typo" in manual and DB=2/DB=1 mean the lowest possible settings, not 6us/3us.

2. Servo resolution test. It shows again 6us for factory default servo. User can improve it to 3us if he buys Hitec servo programmer and changes dead-band setting to DB=1. Even then 3us is far from advertised 1us. When I contacted MikeMayberry again and informed him about second test results, the answer was:

My thoughts are that the servos are 1-2us out of the box and when reset with the programmer.
This is the first time this has been questioned so if the servos were only capable of a minimum of 6us then their performacne would be nowhere near as good as they are. I would suggest you search here and rcuniverse for your answer since you don't beleive what I say.
Mike.

So according to his instructions I ask gurus to point what was wrong with my tests and my interpretation. Better yet if somebody with proper test equipment could repeat these tests and publish their data.

To publish your comments/data and read the full story please follow the link: http://www.rcgroups.com/forums/showthread.php?t=575183

Geistware

My Feedback: (16)

I prefer Hitec over the other arguably tighter deadband competitor because they function just as well and last twice as long or more. Think of it this way, a servo will have a resolution from 900us to 2100us over 180 degrees. This is 1200us over the range or 6us per degree. How tight do you need it to be? I use a lot of mechanical leverage (1:3). I am comfortable with 2us or 3us of deadband.

AlexanderMikhailov

Let me remind you that my question was not a poll about our personal preferences. I like these servos as well. So what? Servos come with 6us resolution instead of advertised 1us -- I do not feel it was fair business. If I was wrong and my test results were not correct please revise them and let everybody know about my mistake.

Anyway I got your message and mostly agree that mechanical leverage 1:3 helps a lot. I fly sailplanes and use appx. 2:7 for ailerons. Unfortunately mechanial leverage for flaps is appx. 10:9 while flap travel is 95 degrees. BTW I am just curious what Hitec servo has 180 degree travel and true 2us dead-band.

mglavin

My Feedback: (31)

How did you test the resolution, or for that matter deadband? Deadband and resolution are NOT one and the same. YES, the stated deadband @ 1us is incorrect, 2-3us is more likely. Remember these servos utilize 3-pole cored motors they have their limitations. Were you suggesting the deadband was set at <2> with your servos out of the box?

You cannot test deadband or resolution with the HFP-10 Programmer, but you can change the deadband programmed parameter as well as others.

AlexanderMikhailov

Hi Michael,

Yes, out of the box servos have DB width set at <2>. It makes sense from an engineering point of view. There is a lot of users who hate digital servo squeak and blame servo manufacturer if it happens. In many cases this squeak is caused by control signal jitter from Tx/Rx. IMHO it is very difficult to educate every user so increasing dead-band width is a "fool-proof" solution for servo manufacturers.

You are absolutely correct that deadband and resolution are NOT one and the same. I even can draw a hysteresis loop for servo_movement(pulse width) function. Anyway servo resolution can not be higher than dead-band.

Regarding your remark "You cannot test deadband or resolution with the HFP-10 Programmer, but you can change the deadband programmed parameter as well as others". Yes, the programmer just reads and displays a stored DB value. It is not a test. Please find any suitable definition for my check action. Here is an excerpt from HFP-10 manual:

Dead Band
This will allow the width of the neutral point to be changed from 3us - 48us.The dead band width is configured in microseconds or "us". It is the "space" the neutral point takes up. The larger the number, the wider the dead band will be. Large planes and surface vehicles often will "gang" several servos together on one control surface. It is important that these servos have matched dead band widths to avoid having them "fight" each other at the center point.
a. To adjust this feature push the input button and the programmer will search for the settings; X:Y will appear with X being the current setting and Y is adjustable from 1-16. The value of "1" is about 3us and is the smallest dead band width setting available.
b. With the knob, select the Y setting you want and "M" button to change the setting.
c. "One" is the tightest dead band width setting and provides the best centering but will also cause the servo to beep quite a bit, the wider the dead band the less beeping you will get around center.

I highlighted red the most significant info. In other words the array <1>:<16> to 3us:48us gives us 3us per programmable step. Servos come with dead-band value set to <2>. It gives us 2x3us =6us. When I asked MikeMayberry to comment he explained it was a typo and <1>/<2> were just the lowest possible settings, not exactly 3us/6us. Actually I could take it but my curiousity led me to...

test servo resolution. Remember that HFP-10 is not only a digital servo programmer but a servo tester as well. Here is another exerpt from manual:

Servo resolution test
To test the resolution of the servo and see the difference in dead band values between other servos, press the DN/R button. Turn the adjustable knob to set the jitter value from 0us -31us.

In technical terms Hitec HFP-10 "servo resolution test" just distorts the output pulse with 0us-31us user selected jitter. Correct me if I am wrong, but this method estimates dead-band width, not measures resolution. It is not a crime as far as actual servo resolution will be equal or even slightly better.

I have got the following results for different DB settings:

DB=2 (factory default)
0us - silence, no servo movements
1us - silence, no servo movements
2us - silence, no servo movements
3us - silence, no servo movements
4us - silence, no servo movements
5us - silence, no servo movements
6us - silent rhytmic squeak, rhytmic servo movements

The result is 6us dead-band (or servo resolution to some extent) which was measured, not read from the memory. It perfectly correlates with DB value <2> as it was stated 6us in the HFP-10 manual. Maybe not a typo, as MikeMayberry stated? I think that the small lie can cause a big distrust. IMHO it was a reason why the original thread with my published results and MikeMayberry(Hitec) answers was deleted from the forum.

DB=1 (the tightest possible setting)
0us - silence, no servo movements
1us - silence, no servo movements
2us - silence, no servo movements
3us - silent rhytmic squeak, no visible servo movements
4us - silent rhytmic squeak, no visible servo movements
5us - silent rhytmic squeak, no visible servo movements
6us - rhytmic gear noise, rhytmic servo movements

Please notice that the user can achieve 3us dead-band (resolution) only in case if he has a servo programmer. Anyway 3us is far from advertised by Hitec 1us dead-band width. I hope the picture is clear now.

mglavin

My Feedback: (31)

Where did you find this information in the Programmer manual? I checked my manual and the online manual to no avail. The Programmer does not actually test servos other than offer a method to drive them and observe the performance thereof. In other words utilizing the servo test function weather "manually" or "auto" the servo responds to drive pulse and sweeps accordingly, this is intended to provide a visual reference to the servos programmed "EPAneuFS" parameters and or fixed as with other brand servos as well note if there are any bumpy or scratchy movement indicating a worn potentiometer and ot the possibly gear-train issues.

Again I'm unsure where you found this feature "servo resolution test"... In order to perform such a test you'd have to dither the drive pulse while observing the feedback loop either with test equipment or physically measuring servo movement or a lack thereof.

How did you ascertain or measure if or when there was/is servo movement? It would be very difficult to step the Programmer consistently in 1us increments. A true test would be to dither the drive pulse in 1us steps and physically note any physical movement IMO (1us = .001" displacement with my test rig), this would allow one to observe the actual servo performance taking into consideration electro/mechanical losses associated with the servo electronics and gear-train.

Pilot232

My Feedback: (3)

Can any of you actually build a model aircraft control system that can move a cortrol surfacr with 1 to 6 miroseconds of servo
input? I doubt it, and furthermore it is physically impossible for a human being to be able to see if that kind of movement
on the model. Lets get real about control movement. The real improvement in todays digital servos is not how fast they are, but in
how they can hold a control surface position.

Travis

exeter_acres

My Feedback: (2)

But Pattern guys Swear they can feel it!!

Geistware

My Feedback: (16)

I guess this is the point I was trying to make.
With the higher refresh rate and the digitals with 3us deadband is electronically stiffer than any analog servo on the market.
There is a greater chance of having slop in your control system that will be beyond anything that we will see at the servo.
While it makes a great academic discussion, it doesn't make any practical difference.

AlexanderMikhailov

S-Test Auto, item i. BTW it is quite easy to make a search on PDF file -- just Ctrl+F.

As I explained previously, HFP-10 tester adds user defined jitter to the output signal. It looks like a sequence of packets (1500us + jitter/2) and (1500us - jitter/2). Packet duration is approximately 0.5 s, so servo arm goes forth and back each 0.5 s IF user set jitter value IS BIGGER THEN servo which is being tested resolution (actually not a resolution but a dead-band).

You are wrong if you think that I manually (and very, very fast) turn the khob right and left with 1us precision -- I am to lazy. I just wonder how could you think of that? User turns the knob to set jitter value from 0 to 31us in 1us steps. Jittered packets are generated by tester electronics.

"No servo movements" means zero rotational movement of servo ouput shaft (servo arm). It was measured with a small mirror glued to the servo arm and a portable laser beam projected by the mirror onto the measuring ruler. Very simple and very precise. And now I think it would be safer to glue the portable laser directly to the servo arm. Anyway -- no servo movements. Nil.

AlexanderMikhailov

Not exactly sir. First of all, not 3us but 6us -- instead of advertised 1us. Would you be happy if your Corvette was running 0-60MPH in 22s? And what would you say if dealer reassured you -- "it doesn't make any practical difference".

Regarding the control surface slop. I do fly European molded sailplanes and make every effort to reduce linkages back-lash. Please follow this link [link=http://www.rcuniverse.com/forum/fb.asp?m=4798474]flutter in aileron and flaps[/link]

Flaps go +5UP/-90DOWN so mechanical transer ratio is 9:10. If servo control impulse range is 1200us and dead-band width is 6us I have only 200 steps thus 0.5 degree per step. Trust me that 0.5 degree for trailing edge make difference for a precisely molded wing.

AlexanderMikhailov

Dear Travis,

It is not about servo speed -- it is about control surface position. Confused why position is measured in microseconds? Please read carefully. If do not understand, read twice. Nothing personal.

dirtybird

My Feedback: (5)

Don't forget that you are testing the servo without a load. Servos behave differently when loaded. Of what use is the servo if you are not going to load it?

mglavin

My Feedback: (31)

Alexander,

I must have read the HFP-10 manual many times to date and never noted the additional note for dead-band comparisons under the auto test header.[&:] I'll have to run this test and look at HFP-10's signal output on my test equipment, did you validate the accuracy of the jittered output?

The laser sounds interesting, is the laser beam diffused in to a line or simply a dot?

I agree that the 5125 specification sheet is miss-leading and obviously incorrect. Hitec is aware of the shortcoming and will remedy same ASAP. However your claim of 6us dead-band of the 5125 seems a bit lofty, I’m going to perform some tests myself to see what I come up.

AlexanderMikhailov

Absolutely. Thanks for your positive input.

To load the servo today I have connected it to the flap linkage. Since the servo neutral point corresponds to flaps 40 degrees DOWN there is a significant elastic reaction from a flap hinge. Linkage leverage is 10:9. The mirror was glued onto the flap surface.

For default DB setting <2> servo started to squeak at 4us jittered signal. Movements were noticed at 5us.
For user selected DB setting <1> servo started to squeak at 2us jittered signal. Movements were noticed at 3us.

I am positive there were no back-lash in the linkage. Moreover it does not matter due to an elastic hinge effect. Maybe it should be better to repeat this test with an old-fashioned spring-return micrometer.

AlexanderMikhailov

So there is a strong case for taking part in conference -- ever smart persons can learn something new.

The laser beam is just a dot but I do not see any reason why it makes any difference in my case. Anyway I hope to hear from you soon regarding your test data (and not only for HS-5125 model).

exeter_acres

My Feedback: (2)

Hey...interesting info...

Can we just go fly now?

AlexanderMikhailov

Since 2004 you have published over 4370 messages. I just wonder if there were something meaningful in them.

akschu

My Feedback: (5)

I have some 5645 servos that don't center as well as I would like. If I got a programmer and turned the deadband to 1, would that make them center better? I only ask because I think this may be applicable to my centering issue.

schu

exeter_acres

My Feedback: (2)

There is a search function... feel free to "research"

Cheers

mglavin

My Feedback: (31)

Yes, lowering the dead-band number to “one” will allow servo to operate its lowest possible dead-band specification. Please remeber this is a sport digital servo with a cored motor, it will NOT center as well as premium coreless digital servo.

Keep in mind there are many variables that affect the centering ability of your servos. system resolution (ATV settings), linkage ratios, linkage slop and more.

What length servo arm and control arm is being utilized? What ATV settings are in play?