AKD Sine Encoder Feedback Interpolation

Greetings,

If we have a encoder which has 10000 sin-cosine cyles per revolution, do you mean that we can have 10000*2^16=655,360,000 distinct counts per revolution? or it is just limited 2^27=134,217,728 distinct counts per revolution? thanks !

Short answer, yes your calculation is correct.

Detailed answer is that AKD drives have two limits to the position word for the number of counts per revolution.

1. All feedback information is scaled to drive native units of 32 binary bits per revolution for rotary motors or 32 bits per magnetic pitch for linear motors. So, no matter how fine the sine-cosine cycles are you can't get more than these 32 bits.

2. The 1 Vp-p interpolator produces 16 binary bits of resolution per sine-cosine cycle and the electronic noise of the interpolator in the drive is specified at 2^-15 of 1 sine-cosine cycle rms.

For the suggested 10000 sine-cosine cycles per revolution you would have 10000*2^16 = 2^29.288 = 655,360,000 distinct position counts per revolution since 29 and a fraction number of bits fits within the scaled 32 bit word.

Please note that the motion sensor input would have to be extraordinarily still to be able to see motion at 2^-29 rev level. Just walking across the floor of an ordinary room with a disabled/not running motor on a table would make shaft motion much larger than 2^-29 rev. You will need a very stable/locked shaft and very quiet analog signals to see the 2^-29 rev level.

For example, a standard AKM 21C motor with DA feedback has a precise optical encoder with 512 sine-cosine cycles per revolution. With the motor on a table, cushioned from vibration, no one walking, no machines making vibration, (late in day when nearly everyone is gone) and the drive disabled if you take a 100 mSec of data in Workbench you will measure the true system noise. The specifications predict <= 2^-9*2^-15 = 2^-24 Rev rms of noise with a perfectly noiseless 1 Vp-p analog source and will have 2^9*2^16 = 2^25 distinct counts per revolution. Just doing this experiment right now I measured 170 32 bit counts of AC rms noise. This level corresponds to 2^7.5 32 bit counts or 2^-24.5 Rev rms noise level which is 1/2 bit better than specification. However, just walking in the room can double and triple this noise level.

So, for the 10,000 sine-cosine cycles per rev you will be 20 times more sensitive to vibrations causing real motion that looks like noise on the reading. It will be hard to directly measure this very small signal without great mechanical isolation.