Hi,

I am designing a 3-phase voltage source using 3x AD9834 chips that are on individual PCB's. They are all individually clocked on their own PCB's using a 1MHz Crystal Oscillator. (So not sharing a clock source)

When looking at the outputs, the frequencies are all exactly 50Hz (In the order of 50.000 accuracy), which is great.

My problem is if I for example, look at two phases together on an oscilloscope, they move away from each other and accumulate a continuously increasing phase error (approximately 0.036 Degrees per second) which is very considerate and a big problem for my application.

Why would this be happening? Are the AD9834 IC's known to have this error? Is it due to them not running off of the same oscillator?

I assume it is the AD9834's themselves leading to this problem but I am not sure how to fix it - I will try and run them off the same clock as a last resort but its tricky in my current situation of surface mount accessibility.

Any ideas would be greatly appreciated! Thanks.

Nick - this is almost certainly a result of using different Reference Clock sources. From a black box perspective, you can look at a DDS as a simple frequency divider. Even though it appears the output frequencies are the same to 50.000 Hz, differences smaller than 5 significant digits will sum up over time and look like phase drift. If my calculations are correct, the a 0.036 degree variation translates to 100ppm difference in average frequency on a 1MHz oscillator. Any variation between your oscillator frequencies will result in a frequency variance between your two AD9834's of the same magnitude. In your case, that difference is negligible enough that the frequencies look the same to the MHz level, and are only observable as a slow phase drift.

A product like the AD9959 (4 channel DDS), which is capable of running much faster than you need and will likely be a more expensive approach, would simplify this because all onboard DDS devices are running from the same reference source.