Falcon is for Fast Laser CONtrol. The main goal is to create a full laser show equipment including galvos, drivers, DAC, software.
|First of all, a small benchmark is necessary to set realistic targets.|
I've gathered all the available data about most well known galvos.
A question of speed
|First of all, speed is really what is needed. It can be appreciated with the small angle deviation time. This is a consequence of 2 physical laws.|
|First, this is en electrical motor. Basically the magnetically polarized rotor is attracted by a magnetic field created by a coil.|
The magnetic field takes some times to establish. This is due to the inductance L of the coil.
The time constant (rise time) is proportional to L/R, R being the coil resistance.
L/R of the best galvos is very near 25 microseconds (L=100”H, R=4Ohms). This can only be achieved with air-wound coils. Using magnetic core increase magnetic field, but L too. You can see it on the table by looking to the G120D. This is the only galvo having a torsion bar to balance the magnetic torque. The use of a torsion bar increase the needed power, that's why a strong magnetic field is needed. Then the use of a magnetic core increase a lot the inductance and leads to 500”s electrical constant.
Nutfield technology achieve the same electrical constant, so the same L/R by decreasing the L and R by a factor of 5. But the torque constant (Nm par Amps in the coil) is twice lower (0.0011Nm/A instead of 0.0025Nm/A).
increasing coil turns increase the torque linearly, but increase L#turnČ
|Inertia can be compared to the inductance. A mass is reluctant to turn around an axis. This reluctance grows with the mass but at the square of the distance from the rotation axis.|
That's why the rotor has preferably larger length than diameter.
Torque to inertia ratio
|The best galvanometer is the best torque to inertia ratio. |
I still have some questions about how to calculate the coil/magnet matching. I've made my calculations much more based on a reverse engineering of existing galvos.
Increasing torque can be done by increasing magnet volume. Magnet volume is preferably increased through the length than through the diameter.
Increasing torque is possible by increasing coils turns. But the inductance is increased and then rise time.
What is the law to match coil to magnet ?
First setup : free magnet
|This is the setup used by Nutfield technology. But this is claimed by General Scanning to cause some unwanted resonances.|
Chosen setup : end bearings
|I have chosen end bearings configuration instead.|
The bearing are two 682HZZ bearings :
- 2 mm ID
- 5 mm OD
- 2.3 mm thickness
|Then we can sort mechanical parts by inertia :|
- Roughly half of the inertia is due to the magnet.
- Then 1/4th is due to bearings. Some other design like WideMove galvos are employing bushing instead of bearings. This saves a lot of inertia but what's about lifetime ?
- The shaft is about 15% of inertia. This can be reduced a lot by lowering the diameter, but the stiffness is too low and then unwanted vibrations can occur and destroy bearing tracks.
- a few percent of total inertia for the sensor
The rotor : first design
|I started with a first design of the rotor. At the beginning I could only find square section magnets. The cylindrical magnets available in the internet shops like www.supermagnet.de have their poles axially on the 2 discs, not radially.|
That's why I've laser cut a sleeve, glued the magnet inside and underwater abrade the difference between parallelepiped and cylinder.
Then I glued the 2 shafts to the sleeve.
The rotor : final design
|Oops... quite the final because there has been an error with the magnetic direction as can be seen on the photo below !|
The final final rotor design :
|Good magnets this time...|
|In this photo, the cylinder is in stainless steel. This was for simulating the setup while waiting for the second magnet shipment.|
|The final setup before assembly. The motor diameter is 10mm and length about 30mm|
|I've determined the coil through a 6 sigma Design Of Experiments. The optimum results have been obtained with xx turns of xx diameter Polyurethane coated copper wire.|
The position sensor
|Last but not least, the position sensor is really a big part of the job. I've formerly played with capacitive sensors and already knew that it's not for DIY.|
I quickly found that optical sensors would be nice and most of the patents are about this way of measuring position.
The position sensor is expected :
- not to raise the inertia
- having good signal/moise ratio
- having good immunity to variable magnetic field
That's why as usual I tried to innovate by choosing the most risky solution : magnetic sensor.
My assumptions is that the field created by the coil can be neglected compared to the field of a NdFeB magnet. Especially the field strength decrease quickly with the distance from the coil.
Thus I've chosen a magnetic sensor close to a second magnet held by the rear shaft. Hall effect sensors are sensitive to distance. I've chosen a Magneto-resistive sensor bridge sensor which :
- must be saturated to work
- isn't sensitive to field strength
- is only sensitive to field direction
- has good resolution (0.05°)
- has big strike (+/-90°)
- has built in amplifier, not to handle ”volts...
The datasheet of this magical component is here :
Still to be fixed
|Here is a list of problems still to be fixed !|
05/28/07 : sensor has been tested and works fine except small non linearity due to the sinus output which is very close (3% error) to the linear output in the +/-20° angle range.
- testing the sensor.
05/28/07 : the setup now integrate a spring which cancels all unwanted vibrations, and without introducing any friction.
- bearings preload. Without any preload a bearing lifetime drop quickly. First, if there is less than 3 balls in contact, there might be some unwanted degrees of freedom, allowing some vibrations to take place. Second, when preloaded, the radial forces due to acceleration are transmitted by all the balls and the contact pressure is acceptable.