[Elecraft] K3 - NR performance vs SGC ADSP2...still no comparison

[Sverre Holm]

> Unfortunately, my K3 NR and NB are poor second performers compared to one
> standard, the SGC ADSP2. I've made my concerns to Lyle and Wayne and hope
> improvement will be placed high on the to-do list using the ADSP2 as a
> standard.

I think the difference in performance may be due to the use of quite different noise reduction algorithms. 

SGC are using what they call Spectral Noise Subtraction (SNS™). I haven't found any more info on that, but it may be a variant of the method published by Berouti, Schwartz, Makhoul,  “Enhancement of speech corrupted by acoustic noise”.  Proc. IEEE In. Conf. on Acoustic, Speech, and Signal Processing, 1979.

The K2 and K3 use a variant of an adaptive filter, I think it is a Normalized Least Mean Square (nLMS) filter with a leakage coefficient.

The spectral subtraction method often requires more computer power than LMS-methods. For more details, I am sure Lyle KK7P can elaborate.


73

Sverre
LA3ZA

Unofficial Guide to K2 Modifications http://www.qslnet.de/member/la3za/K2/mod.html