Why should we use only dyadic length signal in discrete wavelet transform?

Why should the signal length be dyadic while taking discrete wavelet transform? What happens if the signal length is not dyadic. I am working on compressing real time sensor signals containing 2000 samples in 20ms duration.Can some one help me with valuable answers. Many thanks in advance! my mail id is [e-mail]

[Qudus Thanni]

You need to have your data in the nearest power of two for it to have a dyadic length