1/f noise with a passive filter.

around

ut the

ive

but

If your resistor isn't a carbon based one, there is no good explanation in the resistor for your increased noise. There can be explanations in the joints between parts if they are not soldered. In such cases, it us usually worth checking that your test setup doesn't report the same noise from just a 50 ohm resistor or a long coax with a 50 ohm load.

Bad connections can sometimes rectify RF. You may be seeing the local rock and roll station as noise.

Carbon based resistors can have two problems. One is that they can be slightly nonlinear the other is that they can make noise when current flows through them.

Could be that your measurement setup has the 1/f noise. Nearly all frequency sources and spectrum analyzers do.

What happens is you repeat the experiment but replace the L with a short? At 5 KHz, 4uH *is* a short!

John

Done.

All this is in an RF box.

I've already seen 1/f noise with carbon resistors with a DC flow in them, but that was in low frequency. Here I am demodulating at 6 MHz, so this potential noise will be put around 6 MHz (if it goes through the mixer), and I'm measuring low frequencies.

Jean-Pierre Coulon

Carbon resistors exhibit 1/f conductivity fluctuations, and i=v/g. With DC excitation, conductivity fluctuations turn into 1/f noise, and with AC excitation, they turn into 1/f sidebands, which seems to be what you're seeing in your phase detector. Carbon and thick-film resistors both show this behaviour.

Cheers,

Phil Hobbs

But he's wound 4 uH worth of L around a 1M resistor, and he's seeing the 1/f corner at 5 KHz.

I'm thinking it's an instrumentation issue.

John

Hmm, I missed the 1M number, it's true--I was thinking Q=1. The inductive reactance is around 150 ohms, about 10**4 times below R. It would depend how far down the noise floor of the phase measurement actually is.

It could still be the resistor, but it might also be LO noise or even AM noise. It would be worth cranking up the amplitude to make the mixer clip hard, and then looking at the AM noise again.

Cheers,

Phil Hobbs

Our new Aeroflex spectrum analyzer is rated 9KHz to 3 GHz. Actually it goes down to zero frequency, but they rate it at 9K because below there the noise explodes.

John

Am I missing something? X should be in milliohms at 5kHz.

yes you are, the original spec at the start of the thread stated 6 Mhz which works out to 150.72 which gives reason as to why 150 was stated here.

The 5 kHz number isn't what's relevant. The intermodulation is happening at 6 MHz--it's just the conductance fluctuations that are at baseband. Think of an old-time amplitude modulator.

Cheers,

Phil Hobbs

Hi all,

"John Larkin" wrote in = message news: snipped-for-privacy@4ax.com...

In

doesn't

with

local

be

current

MHz, so=20

mixer),=20

With=20

I'm assuming that the two signals into the mixer are in phase quadrature. In that case, the splitter / 2 unequal delay paths / mixer form a 'delay-line discriminator'. As such, it will detect FM noise=20 from the 6 MHz source. To measure really low phase noise of the filter-under-test will require low FM noise from the source -- probably a synthesizer with very low phase noise close-in. In other words, you could be measuring the FM noise of your source.

To test this you could -- Redo the test with a really low noise souce, such as HP 8662. Redo with your present source, but replace the filter with a coax delay line of delay equal to the filter delay.

Also, you say the mixer is "(diode-mixer like)". What does that mean? It is not a diode double balanced mixer? Could the mixer be the source = of noise? You could try replacing the mixer with an actual diode double balanced = mixer.

(I would be very surprised to see any phase noise from a passive L - C filter at 5 kHz offset. At 1 to 10 Hz maybe, due to various mechanical effects, but not at 5 kHz.)

--=20 Regards, Howard snipped-for-privacy@ix.netcom.com

Both are done.

It's a Minicircuit MCL RPD-1, called "phase detector". The internal topology is the same a that of a mixer. Simply the output impedance is 500 ohms, due to unusual transformer ratios. This provides a higher max output DC level than a mixer.

What material would be more professional than a 1-meg resistor to provide a core with a relative permeability of 1? Are there commercial cores for this?

Jean-Pierre Coulon (here "cacas.pam" is what others call "nospam")

around

ut the

ive

but

Another thought:

Do you have a high Q resonant peak in the system at a harmonic of the working frequency? This can lead to there being a lot of the harmonic in the signal. This messes with the ability to measure phase correctly.

I expect that the diodes in your mixer exhibit some 1/f noise and the front end of your spectrum analyzer also does. See the manuals.

Try wood, plastic, glass or ceramic in solid rod or tubular form of some convenient cross section.