[Fwd: LF: Rcvr Sensitivy tests]

[Andre' Kesteloot]

Andy Talbot wrote:

> >
> >> Basically, we used a HP generator tuned to 136 kHz, connected it to the
> >> receiver under test,  and reduced the generator's output until we could
> >> no longer hear the signal. Not a very scientific test, but good enought
> >> to help us select the receivers we will take with us for the LF
> >> receiving test at Nags Head on 15-18 January
> >
> >> The minimum output power we could reliably measure was -127 dBm into 50
> >> ohms (calibrated against another HP spectrum analyzer)
> >
> >> Rycom 3121 (transistorized)  -106
> >> Icom IC-735    -116
> >> Rycom 2174A (tube type)     -116
> >> Icom IC-720    -126
> >> Cubic (US Navy rcvr)    > -127
> >> NRD 525    > -127 dbm
> >> Icom R-71 A    >-127
>
> Some of those are rather poor !!!
>
> To add another Rx to the list ..
> Racal RA1792       -135 dBm ( but with a different pair of ears listening
> for 'not audible'  )
>
> Calibration was with both HP Vector voltmeter and HP scope to double check
> (+/-  0.2 dB).   Signal source was a portable DDS  based module operated off
> batteries.  Any attempt to use a mains supply for this, or an Ailtech signal
> generator, caused signal leakage of around -100dBm into the Rx via ground
> loops etc.
>
> This leakage is a potential route for QRM on these bands.
> The same leakage problem is not evident when frequency rises into the MHz
> region and falls well below audibility.   Has anyone noticed similar effects
> on other receivers or have I got a grounding problem on the RA1792.
>
> Since it was difficult to ascertain when the signal was   "no longer
> audible"   the S/N of  this -135 dBm signal was measured using Dopplergram
> in on it's widest setting.   (Gram is equally suitable when used in spectrum
> analyser mode)
>
>  27dB S/N in a 0.7 Hz bandwidth which equates to 5 dB S/N in 100 Hz - and
> seems about correct for listening to a CW tone.  Rx noise figure is 14 dB
> based on these calculations, which is roughly what it should be according to
> spec and taking into account the error in calculating S/N from a spectrum
> display.
>
> Andy  G4JNT / G8IMR