Newsgroups: comp.robotics
Path: brunix!sgiblab!swrinde!cs.utexas.edu!uunet!noc.near.net!das-news.harvard.edu!honeydew.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!cantaloupe.srv.cs.cmu.edu!deadslug
From: deadslug+@B.GP.CS.CMU.EDU (Richard Voyles)
Subject: F/T sensor replies -- LONG
Message-ID: <CEqJsH.D43.2@cs.cmu.edu>
Originator: deadslug@B.GP.CS.CMU.EDU
Sender: news@cs.cmu.edu (Usenet News System)
Nntp-Posting-Host: b.gp.cs.cmu.edu
Organization: School of Computer Science, Carnegie Mellon
Date: Mon, 11 Oct 1993 13:50:38 GMT
Lines: 333


These are the responses from my request for information on force/torque
sensors (names deleted to protect the innocent). Thanks to all who 
responded. It's not quite enough to make a decision, but it's better
than nothing.

There seems to be some level of frustration with the quality of F/T
sensors available. A few groups have turned to building their own
(Stanford has done this and the Hungarian Academy of Sciences --
see the messages at the end).

I received brief comments on:

	ATI (Assurance Technologies, formerly Lord)
	JR3
	California Cybernetics
	Hungarian Academy of Sciences (long description of products)

Many of the comments are second-hand so use at your own risk.

BTW, I have used the old Lord sensors (not ATI) on several robots
and found the data interface cumbersome. I found it difficult to
control it with the parallel port alone; a serial port connection
was generally required to ensure trouble-free synchronization of the
parallel port. Also, the old Lord manual was terrible. *HOWEVER*, ATI
was gracious enough to send me their new manual and it is much 
improved. The ATI manual is clear and precise, yet succinct. (The
ATI salesman claimed that Lord payed a tech writer by the word, which
is why the old manual was so verbose and obtuse.) Also, they appear
to have corrected the parallel/serial combination by giving the 
parallel port equivalent capabilities to the serial port. And ATI
has gone to silicon strain gages, so they seem to have a much-improved
product.

Nonetheless, we decided to give the "new guys" a try. We are awaiting
delivery of a sensor from California Cybernetics. (They offer a 
choice of color -- how could we refuse?!!  :-)


=========================================================
=========================================================


We use the ATI sensors for our tele-operator work. They're not great but
they are useable and in our opinion probably the best of a bad lot. You
can buy them with or without a controller box. We use the controller box
(which samples at 2kHz), but if you're using a VME crate its probably not
necessary (we use PC's - dont ask why!), the output from the
sensor is a series of single ended analog lines, and each sensor comes
calibrated (i.e., they supply a calibration matrix) so you should be able to
use an A-to-D board to read the signals then multiply the readings by the
calibration matrix to get forces/torques.  This would be advantageous
since you could put your anti-aliasing filters on the analog signals 
(we have to do it digitally). A good Xycom A/D board or equivalent and 
you could probably achieve a 3-4kHz sampling rate assuming you have 
the CPU power.

In my xperience the ATI sensors are very stable - we've done horrible things 
to them. It is occasionally necessary to re-zero them but that's not a 
real hassel, just a matter of taking an offest reading.

The people at ATI are also quite helpful.

******************************************************************

I am using (well soon going to start using 'cos its still being fitted to the
manipulator) an ATI sensor. We orignally were going to use a JR3 but heard from
other research groups that they suffered badly from hysterisis, also ATI's
sales literature seemed to promise a better design (silicon strain gauges etc).

As I say I've yet to use it in anger, but it does seem a neat bit of equipment.
User friendly and all that. The after sales support is really good, they
answered all my technical questions really promptly and competently (pretty
unusual eh ?). Our sensor also has had to have some modifications (for
underwater use), which ATI did very professionally. It actually arrived 2 days
late as they found some problems with the sensor after water-proofing, and had
to correct it and test it all over again - I'm pretty impressed with their
thoroughness.

I'dont know about VME bus connections - the ATI sensor outputs serially (RS232)
and optional analogue and parallel digital output boards are available (which I
have). I don't have the manual with me at the moment but could find out the 
data rate, it does depend upon what format the data is in, method of output, if
limit monitoring is enabled and all sorts of other things.

As for robustness - I don't know yet until mine gets a big bash. It would have
to be big though as it is quite a heavy duty device and exceeding the operating
limits would involve a big bash.

*********************************************************************

Talk to the folks at JR3.  I went to graduate school with John Ramming (one
of the principals); he is a very bright, experienced engineer.

> Also, what about robustness? The old Lord sensors get 
> de-calibrated every time it gets a good whack. (Yes, I know, it 
> shouldn't happen, but...) 

It is very difficult to get adequate sensitivity in all DOFS w/o generating
an overly "stiff" stiffness matrix and have large overload capabilities.
Remember that when robots run into things the forces tend to be very high
due to the lack of compliance.  If this persists in being a problem, a
ball-spring-detent mechanical fuse with limit switches wired into the
E-stops is a very good idea.  This was part of the adaptive robotic welding
cell I helped build at FMC ~8 years ago.  This part gave us a 6 DOF
mechanical fuse; really cut down on broken parts during development.

******************************************************************

Subject: Re: California Cybernetics



Have you had any problems with the force/torque sensors?

	No major problem.

How has California Cybernetics responded to those problems?

	We had some problems during the installation. Zoltan came
	over a couple of times. We can say he responsed very quick
	and the problems were fixed properly.

Have you used any other force/torque sensors before?

	We had Lord F/T sensor before.

What are you using them for and what are they attached to?

	We attached F/T sensor to PUMA 560. We used them
	for force control and dual-arm coordination control.
	The detail can be found in our paper in IEEE R&A Conf.
	1993 at Atlanta.

Is there anybody you know that has used both the CCC F/T sensor
and a sensor from ATI (formerly Lord Corp.)?

	I know that Johnson space center bought CCC F/T sensor
	
What is your general opinion of the sensor and the company?

	Sensor is good and Zoltan is a nice guy. Everytime I need
	some help, he always try his best.

Are you using the processor card from CCC or are you interfacing it
yourself?

	We used the processor card from CCC. 

What bus architecture are you using to process the data?

	We have multi-bus Controllers. I believe JPL
	used for VME bus controller.

We replaced Lord F/T by CCC F/T. The basic reasons are

	1) CCC F/T can give us high sampling rate (up to 1000Hz).

	2) It is easy to interface with our controller(We have CCC 
	   Controller).

**********************************************************************


We've been using ATI from a serial port of a PC. I'm also in the market for
new sensors shortly and would like to suggest we exchange any info collected.
I'm a physiologist using them for limb forces : more sensitive the better
and 100-200Hz rates acceptable. The ATI FT210 was pretty good, but I dont
think they still sell it. We tried JR3 but had enormous RF noise problems 
and ended up returning the one we tried. I know JR3 is well regarded in 
Japan so maybed things have changed...


***************************************************************************

Referring to your 6-axis Force/Torque sensing related e-mail message this is to
send you short form information on our 6-axis F/T sensor devices. Our research
team at the Computer and Automation Research Institute, Hungarian Academy of
Sciences  has several years of experience in designing 6-axis Force/Torque
(F/T) intelligent robot sensor systems. Our MiniForce system has been commerci-
ally available for two years. According to its speed and functionality the
MiniForce is very similar to the announced new ATI sensor. The MiniForce is
based on the 68070 microcontroller and provides 1kSPS maximum sampling rate.
The MiniForce was primarily designed for F/T feedback based supervisory
control.

The very new NetForce system will be released (hopefully) at the end of this
year. The NetForce is based on the ADSP-2111 digital signal processor and
provides 12.5 kSPS sampling (and output) rate. The Signal Processing part of
the NetForce system was primarily designed for acting as a high performance
slave sensor device of an open robot controller. Based on its compact design
the unit can be directly mounted on the robot arm. The configurable,
bus-oriented 16-bit wide Host Interface Port can be easily connected to
standard bus architectures like VME, ISA, E-ISA etc. The synchron Serial Port,
among others, supports sensor level network in a multi-robot application.
The integrated F/T functions extensively support guarded robot motions and
active compliance tasks. For more detailed information on the NetForce system
please see "Digital Signal Processor based Force/Torque Sensing for Advanced
Robot Control" Proc. of the Int. Conf. on Signal Processing Applications and
Technology (ICSPAT), Cambridge, MA, 1992, pp. 671-677.

As a part of an academic institution we are seeking co-operating facilities
with other research teams in the field robot sensing (F/T and Vision). If
you are interested in working with the NetForce sensor we can establish a
joint project where you can use and test the sensor and based on your
experiences we can improve (if needed) the functionality of the device.
The same method worked very well in case of the MiniForce, where a highly
skilled "tiger team" tested the device in Finland.

Upon request I can send you detailed materials on the topic.

In case of any comments and questions please do not hesitate to contact me
immediately.


Sincerely Yours:


Tamas Ledeczi


Computer and Automation Research Institute
Hungarian Academy of Sciences
H-1518 Budapest, Kende u. 13-17.
Hungary

Phone:  (+36-1) 1811-760
Fax:    (+36-1) 1667-503
E-mail: h1183led@ella.hu



Both the MiniForce and the NetForce system contain three parts: the 6-axis F/T
transducer, the Signal Processing Unit (SPU) and the Robot Interface Unit
(RIU). All the real-time F/T measurement and preprocessing tasks are carried
out by the SPU. The SPU can be used as a stand-alone device if the robot
controller has improved communication facilities. In case of closed controller
architecture where only the usual binary I/O lines are available for external
communication and sensory feedback, the RIU can be used as a communication
server between the SPU and the robot. The main features of both sensor systems
are as follows:

The SCT-04 six-axis F/T head family (product of the TARA Ltd., Hungary):

Nonlinearity:                   0.1% of the rated output
Hystereiss:                     0.1% of the rated output
Repeatability:                  0.1% of the rated output
Allowable overload:             300% of the rated output
Op. temperature range:          -10..60C
Cross-sensitivity:              max 1% with arithmetic correction
                                max 5% without any correction
Mass:                           0.4 Kg

Rated force  (Fx, Fy, Fz):      100, 200 or 500N
Rated torque (Tx, Ty, Tz):      20 or 40Nm

In case of static overload within the allowable limit the manufacturer
guarantees that there is no need for de-calibration. I will contact
the designer for further information if you need more details.

MiniForce

SPU:    - 8..12V DC supply;
        - robot mountable;
        - six filtered differential analog channels with 12 bit resolution;
        - SCC68070 core processor;
        - RS422 or RS423 (user selectable) asynchron serial port with max.
          38.400 baud transfer rate;
        - internal or external sampling 'clock';
        - 128 byte on-board non-volatile storage;
        - well defined Interprocess Communication Protocol provides full
          programability;
        - variety of built in services support guarded motions, real-time path
          corrections, tool tranformations, zero point definitions etc.

RIU:    - SCC68070 core processor;
        - opto-coupled binary I/O port (8 input, 8 output and 4 handshake);
        - TTL binary I/O port (8 input, 8 output and 4 handshake);
        - RS-422 asynchron serial port for SPU communication;
        - terminal connection with menu-driven programming frame for F/T
          application developement;
        - high capacity EEPROM storage for user defined data structures;
        - well defined IPC.

The price of the MiniForce depends on the required configuration:

        F/T head + SPU          :               4500 USD
        F/T head + SPU + RIU    :               6500 USD

To support our customers final decision we offer one month test period
(free of charge) for testing the MiniForce system in the required application.
We also offer one year warranty and further software engineering if you need
new F/T services to be integrated.

For more detailed information on the integrated F/T services please see,
'Intelligent Force/Torque Sensing Functions Enable Easy Interface to Robots'
Proc. of the 11th Int. Conf. on Assembly Automation, Dearborn, Michigan, 1990.
SME Technical Paper Index: MS90-818.

Delivery deadline is 6 weeks from order.

NetForce

SPU:    - 8..12V DC supply;
        - robot mountable;
        - 12.5 kSPS sampling rate;
        - six filtered differential analog channels with 12 bit resolution;
        - ADSP-2111 DSP core processor;
        - configurable, bus-oriented, 8/16 bit wide Host Interface Port;
        - network-oriented synchron serial port;
        - 2Kbyte on-board non-volatile storage;
        - well defined Interprocess Communication Protocol provides full
          programability;
        - variety of built in services support guarded motions and active
          compliance tasks in both supervisory and servo feedback level.

RIU:    - ADSP-2111 core processor;
        - opto-coupled binary I/O port (8 input, 8 output and 4 handshake);
        - 6 bipolar analog outputs for the six F/T axis;
        - network-oriented synchron serial port;
        - terminal connection with menu-driven programming frame for F/T
          application developement;
        - high capacity EEPROM storage for user defined data structures;
        - well defined IPC.


Since the NetForce has not been released yet, there is no final price.
If your team has any interest in testing the prototype of the NetForce
in the frame of a cooperation, please send me brief information on your
department's fields of activities, results etc. Thanks in advance.

*********************************************************************
*************************************************************************
