[Radio Wrangler]

If you're really testing a receiver, say a new design, one thing you have to do is to tune it to a chosen frequency and then sweep a strong input from DC to daylight and verufy that only the wanted response is strong, and that all spurious responses are suppressed within spec.

So your receiver will operate with signals down to the -100dBm and lower range. A big bad interferer might be -20dBm and you might have to get spurious responses to out of band signals down by 60 or 80dB. So you dial up -20dBm on the sig gen, get the 'phones on and slowly screw the frequency up. You'll find images, you'll find spurii due to synthesiser ref freq sidebands and all that sort of stuff. So you have to check that the strength of these responses is less than that to a -80 or -100 dBm signal. But the big ones you hit are tha harmonics of your sig gen, when the generator tunes across an integer submultiple of the frequency the receiver is sat on.

Oh rats! With only 30dB harmonic suppression in the sig gen the response you get is more like that from a -50dBm signal. 50dB more than you're listening for.

Now you can't just say 'Oh, it's only the sig gen' and go on, ignoring it. How do you know your receiver front end doesn't have a harmonic generation problem all of its own?

So you have to build a low pass filter to remove the harmonics down to a low enough level to prove the receiver. If the receiver tunes over much range, that filter expands to a group of them.

Yes, sig gen harmonics are very poor compared to what is sometimes needed.

Anothe failing of modern sig gens is broadband noise floor. The 8640, 2008 and all that dort of thing have cascaded wideband amplifiers and a wideband output match... resulting in a wideband noise contribution. This gets really nasty when you use it as an LO and the signal is wideband, or there are many signals you get a noise X noise mix dropping the whole lot into your receiver bandwidth. Nasty. THe matest sig gens with I/Q fncy modulation have to go to rather low levels for modulation, then lots of gain.... so lots more broadband noise.

THe old Marconi TF144H was fairly low on broadband noise, being a single valve power oscillator with attenuation. The HP606/608 had tuned buffer and output stages tuned with extra coils in the turret and extra sections on the tuning capacitors.. Their broadband noise was amongst the best.

So it isn't all just phase noise.

David