small rotary sensor for a very small walking robot

small rotary sensor

I am doing a research in microrobotics, and it is a walking robot
(size is about 4cm x 4cm x 3cm).

I have made a good progress for a year now, and I am redesigning a leg.

I need to have a rotary sensor (non-contact preferred) that can be as small as
3mm x 3mm. The sensor will be on the knee to measure the angle of the leg flexion.

I am thinking of HMC1501 (4mmx5mm) which measures the direction of the magnetic field. So I will have a magnet also that rotates with the leg.

But I would like a smaller version. Any suggestions?

Thanks.

rotary sensor tutorial

Make 3 identical Ø3mm PCB's with hole for "knee" shaft and encoder patern on remainder of PCB's surface. They should be as thin as possible. Put all 3 together, one in center should rotate with one side of leg and other two with other side. two PCB's that are moving together should be rotated 90º apart. What you get is capacitive resolver-encoder. You put AC sinewave excitation to center PCB and read SIN and COS signals of the other two.
Board thickness and trace resolution will be dependant on your PCB manufacturer and that will dictate excitation frequency and resolution of angle measurement.
Reading of this sensor is same as "sinewave encoder" where each encoder step is broken to desired number of resolver substeps.
As for hall sensor i think smallest is SO-8

kender positioning site:edaboard.com

Hello,

I understand, that an absolute position sensor (with limited resolution) would be more suitable than an incremental encoder, which also needs an additional index sensor and initialisation procedure. Thus the said xy hall sensor doesn't look bad, execpt for it's size. Other options would be differential capacitor or inductance, both could be downsized to micro PCB. I've been using this principles from time to time - in regular "macro" scale. But AC signal processing would be implied in both cases (as with the suggested capacitve encoder, too), thus it's not said that the overall form factor would be smaller than hall solution.

One dimensional analog hall sensors are available in smaller case, but would need more complex calibration and offer less stability than xy sensor.

Regards,
Frank

At the risk of being obvious - a potentiometer could be your sensor. You get a single signal for angle, instead of cos and sin. On the down side, the pot would be a contact sensor.

By the way, I'd like to invite your question to a yahoo group dedicated to sensors: **broken link removed**

As a side note - human body doesn't have rotation sensors. Instead, each muscle has a displacement sensor. If you replace a "knee" rotation sensor with a displacement sensor in a "thigh", would that have any advantages?

Thanks for your kind replies.

After reading your suggestions, I am interested in capacitive encoder. I have never used one before. Is there a good example or tutorial regarding construction and operation? (It does not look too complicated to me, though)

I have considered the potentiometer option, but I could not find a potentiometer small enough and still have a good life (10,000+ cycles).

Displacement sensor is definitely welcome, but I do not know of any suitable ones. This also has to be small. Do you have any suggestions?

Thank you very much.

Both FvM and Kender had a good point.
As for capacitive resolver, sensor itself is small but would require extensive signal conditioning. As FvM pointed out, to avoid encoder counter nulling procedures, you should have resolver configuration. That is single surface that would cover 180º and end PCB's are rotated 180º in respect to each other.
Signal conditioning should be integrated so it would simplify circuitry and keep dimensions small. There are resolver to digital converters IC's that you might be able to use (e.g. AD2S1200). There are capacitance to digital converters that could be used for this purpose as well, where MCU would do all position calculation from raw capacitance data.

On the other hand Kender had very good point for "thigh sensor" which could possibly be done simpler and smaller. Guitar string or fish line could pull aperture located on the main body when knee is flexed. Aperture could interupt light beam in small fork opto isolator and it can also have opening shaped in a such way to produce linear output to knee angle. This would give you virtually smallest sensor possible at the knee and very easy interfacing to MCU.

FYI, there are also potentiometers that use conductive plastic and are rated for long life and high cycle rate. (20e6+ cycles)

MathGeek, if it's not a secret, what actuators are you using for such a small leg?

Sorry for the late reply: email notification was not working.


Sinisha--those are very good points. Also, if you have recommendations for the particular potentiometer type you mentioned, feel free to post them here.


Kender--I am using SMA springs to actuate the legs. They are wonderful except extremely inefficient (<2% efficiency). So with a 1.5g Li-Poly battery, the entire robot should run an hour or so (which is not too bad).

I am looking towards another mechanism to actuate the legs. I am thinking about Dielectric Elastomer right now, which is highly efficient (>80%), but requires several kV. As you might imagine, creating several kilovolts is not trivial especially when electronics is should be extremely small--I posted this question on analog section of EDAboard forum.

Here is an update:

I have created a small plastic robotic joint (~3mmx4mm) and put magnet on one side and put the AMR sensor on the other side.

Works wonderfully, but is too large with the chip and too heavy with the magnet.

I am really looking into the capacitive resolver-encoder method.

Any basic help would be appreciated.
Thanks.

anyone?

Click to expand...

Post your question here: **broken link removed**. Different crowd might come up with different ideas.

I joined the group as recommended.