(Allison et al 2011) detail the method used by the ARRL in their test reports on equipment.
Effectively they calculate NF=-174+27-MDS where MDS is measured
in the CW mode using the 500 Hz, or closest available IF filter (or audio filters where IF filters are not available).
One flaw in this method is that the factor 27dB in the NF formula implies that the
Equivalent Rectangular Bandwidth (ERB) of the receiver when measuring MDS is exactly 500Hz. More correctly, the formula should be NF=-174+10*log(ERB)-MDS. The error could be significant, especially with the
closest available… provision in the test requirement.
For example, above is the measured IF response of a TS-2000 set to 500Hz bandwidth. As filter responses go, it is very good, having quite a flat region, better than most crystal filters used for CW, but the ERB is actually 430Hz, that is 0.66dB less noise than a 500Hz ERB filter and any NF based on 500Hz is low by 0.66dB.
The red plot is that of an idealised filter of the same ERB.
A more fundamental flaw is uncertainty in measured MDS leads to a relatively large uncertainty in NF when NF is small. The test reports do not specify the uncertainty of MDS, it is unlikely (Duffy 2007), (Duffy 2007b) that with receiver bandwidth of 500Hz (as specified for the MDS test) and the HP339A instrument used, that uncertainty to 95% confidence level is as low as ±0.5dB probably closer to ±1dB. The method used is just not suitable to low noise receivers.
Giving NF rounded to 1dB is not very informative for receivers with NF below 5dB, and fairly useless at 2dB as in (Wilson 2012).
- Allison, B; Tracy, , M; Gruber, M. 2011. Test Procedures Manual Rev L. ARRL Newington.
- Duffy, O. 2007. Uncertainty of the noise sampling process. VK1OD.net (offline).
- ———. 2007b. Noise measurement uncertainty calculator. VK1OD.net (offline).
- ———. 2014. ARRL Test Procedures Manual (Rev L) – Noise Floor test. https://owenduffy.net/blog/?p=1165 (accessed 15/03/2014).
- Wilson, M. 2012 ICOM IC-9100
MF/HF/VHF/UHF Transceiver In QST Apr 2012.