Baldor servo with resolver and encoder feedback???

Hi,

I have an issue with an old Baldor servo motor(from a very old Staubli RX90L) and its feedback system.
Yahoo! - 302 Found
Yahoo! - 302 Found
My first impression when i opened the motor was that clearly the feedback device was a resolver but then i realized that the shaft was also connected to a series of gears and a totally apart circuit. It seems to be a kind of encoder or something based on this link: (yes sorry, thats all i got)

FSI International 280309 002 Encoder for RX90 Joints 1 2 3 4 Imas 70 | eBay

More Images:
Yahoo! - 302 Found
Yahoo! - 302 Found
http://flic.kr/p/kD2AGT
http://flic.kr/p/kD4mvo

Video of mechanics of encoder:
https://www.youtube.com/watch?v=fVdE...RVBJgoX-4X6BFg


After a big search it came to me the term: "multispeed resolver" but i'm not sure because this is an old motor(1995-1997?). The two cilinders on top of circuit (optic encoders may be) are connected through the gears to the shaft rotation. Below that circuit and cilinders is the resolver: Tamagawa TS2620N271E14.
Anyone recognize this system ? someone can tell me how it works?? the gears ratios are 1:4 and 1:80(more or less, 1 cycle needs multiple shaft turns) i need more info about the output of this and why is used???? I seems i got 3 signals, one from the tamagawa resolver, one from the reduction 1:4 gear ratio encoder and one from the reduction 1:80 gear ratio encoder. This 3 signals go through the circuit(brown red). HELP!

BTW: Let me know if the images are down. Thanks

It was typical of older motors that used resolvers to use a gear train to increase the resolution.
Al.

looks like the std motor frameless resolver in back of motor unremoved. it was used to commutate the motor and run the current and velocty loop. the dual gadget was used for absolute feedback over many turns of motor shaft (80*4=320 in your case) -but used resolvers not encoders. how do you know those 2 fb things are encoders? from someone's misdirected ebay ad?

we did this for years before absolute multiturn encoders were developed.

multispeed is just name for how many poles a resover is: single speed = 2pole, 2 speed = 4pole etc. no sense in using anyting but 2 pole in dual fdbk PAU (position - adapter?....too late in day to remember... - unit)

Thanks guys for the info!
Mike, using the dual gadget for absolute feedback makes a lot of sense knowing that the motor was used to move a robot joint with limited freedom. Actually i was reading something about that. Thanks a lot. You're right, my source is a ebay ad(that is all i can find searching for:IMAS 70.23), i'm not sure if those are encoders. Looking for Baumer CH-8500 takes me to another thing:

Baumer Electric CH 8500 FHDK 10N3103 Photoelectric Sensor New | eBay

(ebay ad, i know, but the images in the ad shows(in the box) that the article Baumer CH-8500 is in fact (reduced size i guest)a photoelectric sensor.) That clue leaded me to think that the device was somekind of optical encoder, but it was just an assumption.(i'm just a student, my experience is limited). Many links also show a photo electric sensor too.
One think that i dont understand is that when i look to the feedback output in the oscilloscope, it shows the expected wave of a resolver but it takes me many mechanical shaft turns to be able to complete one cycle of the resolver wave so, i think the resolution was reduced somehow, right? That is confusing because the RX90L was a "precise" robot. I was expecting the opposite result.(many resolver cycles for 1 mechanical cycle, like a multispeed resolver).

ur on the right track.... as I said, they are likely resolvers rather than encoders; encoders were typically increment - no absolute data. resolvers are absolte - produce a full sine wave per rev (for single turn units, 2 sinewaves for 2 speed etc).

I think you will find the 4:1 ratio one makes one full sinewave when the input shaft is turned 1/4? or maybe the 4 - too early to think it thru right now.

and the other one of course will make 1 sine wave thru the total 80 turns. So with these two absolute signals, you can begin to envision how it works: one resolver is geared crude so it produces a single absolute sinewave over WHOLE travel of the robot joint, the other then provides the fine data; ie., the single sinewave is divided into 80 more segments but being 80 turns there is no absolute data. so control uses it but also watches the single turn over full travel to see which 1/80th portion of the sine wave it is in. the control only has to do this on power up to find which of the 80revs it is in, then just dump that number of counts into the pos register and begin using the faster turning (80:1) resolver for simple feedback.

accuracy and resolution are two totally different things remember.

rsolution in your case seems to be 80 sinewaves over full travel: isn't that enough??? each sinewave in that era control got devided up into typically 65,536 counts so they had 5,242,880 counts over the joint rotation - what? 270 degrees? so 1 degree was 19,418 counts! or >300 counts/arc min or 5 counts per arc sec.

well, that era resolver had an ACCURACY of typically 30 arc MINUTES and REPEATABILITY of maybe 1 arc minute, so there you go.

If you want to be sure, just count the revs while whatching the output of the resolvers. or to be sure they arre rsolvers, count the wires: ref/sin/cos = 6 total and ref is usually 30-100 ohms, sin/cosine 20 ohms..... encoders are high resistance and need +5v and comm supply too..... lots of clues.

do you need to fix something or just playing with it? those resolvers can be replaces with std size 11 ones if needed....

Mike,
Now i begin to understand what i'm dealing with. Pretty cool by the way.
Actually me and my partners(university lab) are trying to reuse this motor as a project for a mobile device. We're recycling them in order to save some money and of course we need to understand the output to be able to control it. We'are just trying to use the full capabilities of the motor and feedback and get good accuracy.
I was not able to go to the lab(lot of work) so i dont have info about the output of each resolver. I'll try to have it by thursday. Thanks a lot.

Modern systems now in order to increase the usually relatively low res of the resolver etc is to use the arc-tangent function of the two quadrature sine waves.
Al.

Hi guys,

I haven't been able to measure the outputs in the lab but meanwhile i have a few questions:

-Knowing that the feedback system we've been talking above is composed by 3 resolvers:
Yahoo! - 302 Found (image link works)

1-The Tamagawa resolver (the one connected directly to the shaft in the picture).
2- The 1:4 geared resolver.
3- The 1:80 resolver.

Here's the thing:
Why do these devices need this electronics ?? Yahoo! - 302 Found (image link works)
I mean, a resolver is just an inductive analog device. Right? I might thing that the shown circuit is kind of a ADC but the fact is that the 3 signals of each resolver go into the circuit and the ONLY output of this circuit are a SIN and COS signals,i mean, just 4 wires like a single resolver. How is possible that 3 (sin/cos) analog signals are fused into 1 (sin/cos also)analog signal and not treated separately ??whats the point??

Thanks

the tamagawa resolver has nothing at all to do with the dual PAU unit (position analog unit) mounted as option on the motor. the tamagaw is to make the motor work, totally sep from the pau.

often the pau has hall sensors also for end of travel limit switches - perhaps that is function of the pcb?

how many wires coming from the pau? you should be able to see the two sets of 6 wires per resolver and where they go...

not sure beyond that what details are on that device without studying it in detail.