High frequency signal averaged QRS (HFQRS) has previously been analyzed without any definition of optimal passband, processed leads and other settings. We analyzed HFQRS in orthogonal X, Y, Z leads, in 12 frequency windows (bandwidth 100 Hz, middle frequency f0 from 100 to 650 Hz, step 50 Hz) in three groups of subjects: Healthy (N=182), Ischemic heart disease (IHD) (N=237), and Dilated cardiomyopathy (DCM) (N=87). All ECGs were recorded at St Anne’s University Hospital Brno, in supine position, with 5 kHz sampling frequency and dynamic range of 24 bits. Analyzed parameters were HFQRS maximal amplitude (Amax), HFQRS power (PW) and HFQRS fragmentation (FR) based on normalized length of HFQRS line. The aim of the study was to assess differences between groups in relation to passband, lead, and parameter. The reproducibility of parameters (independent analysis of first and second half of measurement) was the secondary aim. Results: Significant differences (p<0.0001) exist between healthy-vs-IHD and healthy-vs-DCM in FR in all passbands and leads, in Amax and PW in lower frequencies, up to f0=400 Hz. Comparing IHD and DCM, the maximal difference (p<0.0001) exists in FR in lead X and f0=150 Hz. In parameters Amax and PW differences exist only in lead Z and f0=100 Hz (p<0.0001). If the group IHD is divided on two groups according to Heart failure (HF) presence (184 positive, 53 negative), the maximal difference (p<0.0001) between the groups is in lead X and f0=250 Hz. Relative irreproducibility increases with f0 from 4% to 28%, is similar among different subject groups, and higher in lead Y and in FR parameter. Conclusion: HFQRS parameters depend significantly on selected passband and this dependency should be reflected in diagnostic applications. The leads X and Z are optimal for FR, and for Amax and PW, respectively.