High-Speed Angular Position Measurement for Precision Applications

[Eidos]

Hello Everyone :)

Do any of you know of a floating point Analogue to Digital converter that can handle mega-hertz sampling frequencies?

I need a really precise angular acceleration transducer which I'll integrate digitally to get angular velocity. I am using an FPGA to do the integration.

Resolution on the order of 10^-3 and frequency 10^6 hertz.

The environmental conditions are -20 to 70 degree C.

Basically, is this sort of thing feasible?

 

[mgb_phys]

Never heard of a floating point ADC
You normally just amplify the input to the full range of the ADC (some have internal scaling amps) and then convert the digital output into whatever units you want in software.
A resolution of 1:1000 ( ie 10bits) at Mhz is easy, over that temperature range you will need some sort of temperature controlled reference.

Can your accelerometer really work at MHz? Sampling much faster than the analogue circuit can respond isn't gaining you anything

 

[Eidos]

Im using intensity modulation of a laser beam on a spinning disc.
In Ascii art.

_________________...... /____ DISC
^...|....^......\
...|.....*
...|.....*
...|.....*
...|.....||^ - reflective strip

|| - laser

I've made a design change for simplicity's sake. I am going to count the pulses that the reflective strips send back. There are 500 equally spaced apart strips on the outer rim of the disc with a radius of 0.0174 m which implies that the circumference is 0.957557441 m.

Which is roughly 1m.

I therefore need sub-millimeter placement of reflective strips to get a resolution of $5\times 10^{-3}$. Now the disc is spinning on the order of 10^3 rad/s which implies that the sampling needs to be on the order of Mhz.

Can you see any problems with this setup? I can't think of any interference or sources of noise if I use monochromatic light.

This scheme is inherently digital so there is no need for the ADC anymore, by the way I did find an ADC that met the spec called ADC291 if anyone is interested in a hardcore ADC chip. But can sample on the order of Giga samples per second.

Its not floating point though, sorry about that only fixed point 8 bit, 2 channel.

 

[MATLABdude]

If you just need position / velocity, you might be saving yourself a whole lot of trouble by getting a shaft encoder (these come in incremental and absolute variants--choose accordingly based on your needs):
http://en.wikipedia.org/wiki/Rotary_encoder

Big fan of these guys for reliable cheap(ish) encoders (all-in-one units that include the disc and IR module)--but if you Google for shaft (or rotary) encoder, you'll find a fair number of hits:
http://www.usdigital.com/

You generally use a pulse counter (for the incremental ones), or get a reading from the absolute one (the outputs of these vary greatly). There are also some nice ICs that do quadrature decoding (so you can figure out which way the disc is turning) and various other motion-control tasks:
http://www.usdigital.com/products/interfaces/ics/

 

[Eidos]

If you just need position / velocity, you might be saving yourself a whole lot of trouble by getting a shaft encoder (these come in incremental and absolute variants--choose accordingly based on your needs):
http://en.wikipedia.org/wiki/Rotary_encoder

Big fan of these guys for reliable cheap(ish) encoders (all-in-one units that include the disc and IR module)--but if you Google for shaft (or rotary) encoder, you'll find a fair number of hits:
http://www.usdigital.com/

You generally use a pulse counter (for the incremental ones), or get a reading from the absolute one (the outputs of these vary greatly). There are also some nice ICs that do quadrature decoding (so you can figure out which way the disc is turning) and various other motion-control tasks:
http://www.usdigital.com/products/interfaces/ics/

Awesome! Thanks for the links :)

Shaft encoding was something that I was looking at but was worried about the extra moment of inertia that it would add to the spinning disc. I forgot to mention the application, the goal is to measure the rpm and angular acceleration of a reaction wheel for a micro-satellite with a view to estimating the angular velocity of the satellite as a whole. The rpm is just extra bang for your buck since it can pop out of Digital Signal Processing by simple modulo arithmetic.

So if I know angular position (which I need to know to within $5\times10^{-3}$) I can numerically differentiate it twice to get the angular acceleration (but this amplifies noise). This is the reason why I wanted angular acceleration so that when I integrate it I don't amplify high frequency noise.

More importantly, the goal is to design a smart transducer. Which means I am not allowed transducers with built in signal conditioning.

Any further help or tips would be very much welcome :D

 

[mgb_phys]

Extra momentum for the shaft encoder is tiny, a bigger worry at high speed is any imbalance - the reflective strips stuck on for the laser might be enough to unbalance it.
Another common technique is either a hole through the shaft for an optical counter or a number of small flats machined on the shaft.

How fast is it spinning? From shear inertia how many rotations would it take for it to vary enough to exceed your requirements? How often would you have to measure position to achieve this.

Shaft encoders are great but high precision ones require a very close tolerance between the encoder and read head - which implies very stiff bearings.

 

[Eidos]

Max is 8000 rpm

I need to measure angular position in the Mhertz range in order to resolve the angular velocity on the order of $$5\times 10^{-3}$$ rad/s .

So far I can't see the reflective surface as unbalancing the reaction wheel, although I will admit that I have not even done an order of magnitude calculation. My intuition tells me that as long as I place the reflective strips symmetrically (which I am going to), all will be well.

A fairly big snag that I stumbled across today, I cannot tell which way the wheel is spinning using the proposed measurement scheme. Do you have any ideas about how I could get that information? A simple binary output that tells me whether it is spinning one or the other will suffice.

 

[mgb_phys]

Max is 8000 rpm
I need to measure angular position in the Mhertz range in order to resolve the angular velocity on the order of $$5\times 10^{-3}$$ rad/s .

So you need to make a measurement every 3degrees

Easiest way is just an angle encoder
All the normal machine tool instrument makers (Heidenhain/Renishaw) make encoders that will do that

So far I can't see the reflective surface as unbalancing the reaction wheel,

8000rpm isn't that fast - I was thinking of high speed optical balanced mirrors.
Most optical encoders do a quadrature scheme which gives direction.
If you are just using markers simply put two together and then one further apart.
"mark short space mark long space" or "mark long space mark short space" tells you the direction.