[time-nuts] LNA and Alias
kb8tq at n1k.org
Mon Dec 22 09:39:23 EST 2014
At the moment Comcast and ko4bb.com don’t seem to like each other. I can’t refer directly to the schematic or the page.
What you *should* have (assuming 10 MHz inputs for clarity):
1) The RF inputs go into the mixer from what ever sources you happen to want to test. At least one source needs to be high level.
2) The mixer output goes to an L/C lowpass filter. That filter serves several purposes:
a) It resistively terminates the mixer at RF (both at 10 and 20 MHz) in the proper IF impedance
b) It passes the phase noise information on to the LNA
c) It rejects all RF going to the LNA
3) The LNA can be just about anything provided:
a) It handles the signal levels without overload
b) It does have low enough noise (that depends a lot on the mixer and sources)
c) It has enough gain
d) Terminates the mixer in a reasonable resistive impedance at audio.
4) The LNA feeds two things:
a) Your FFT box
b) Your DC bias box
It sounds like the FFT part is working so the DC bias may be an issue. The bias box is used to force the two oscillators into quadrature. It forms a PLL around the oscillator pair. With the two signals 90 degrees apart your mixer has a ~ 0V output. That is the point it is most sensitive to phase noise and the least sensitive to AM noise.
There are lots of ways to do this. Since I can’t see the schematic. Here’s one based on an RPD-1 (500 ohm out) mixer:
One side of the mixer is grounded, the other feeds the filter.
500 ohms in series with 820 pf to ground as the input to the filter. Some sort of coil in the vicinity of 100 uH as the first series lowpass element. Next a 470 pf to ground. Then another 100 uH in series. Another 470 pf to ground. Another 100 uH in series. Another 470 pf to ground. At this point you have a three coil and five capacitor lowpass filter. You should poke it into spice to make sure it’s not going to be a problem with your parts. The issue is cutoff at the highest frequency you want to look at phase noise. You may need to tweak values a bit. The final stage may be overkill depending on the quality of your coils. You will always have a tradeoff between highest phase noise frequency and lowest RF frequency with this setup.
LNA can be a good audio op-amp. Run it in positive gain mode. Termination resistance for the filter is simply set with a resistor to ground. I prefer to use 5K for the RPD-1’s. It gives you a bit more output voltage. It also makes the cutoff of the lowpass a bit lower.
The DC bias box is an op amp plus a pot, resistors and capacitors. You need to set the output to the EFC voltage on the OCXO’s. That will vary between different parts. One pot is for centering this up. You need to set the loop gain, so feedback resistors on the op amp need to be adjusted. Some sort of R/C may be used to roll off noise. The cutoff frequency of the loop will determine the lowest phase noise frequency you can check unless you measure loop dynamics and correct all your data.
Now that that’s all working, you need to calibrate the setup. Two common approaches. Both use a beat note formed when the bias box is shut off:
1) Measure an power or voltage at the LNA output and do math based on some assumptions.
2) Capture the full beat note and look at the actual slope as it crosses zero.
You pretty much have to do number 2 before you can use number 1. The LNA needs to have low enough gain in this case to not distort passing the full signal. The math for 2 is pretty simple. Each cycle is 2*PI radians. Phase modulation is normalized to one radian (yes it’s phase … ). You get a radians per volt number and move on.
Wish I knew what Comcast was doing this morning ….
> On Dec 22, 2014, at 4:26 AM, Loïc <loic.moreau at eai.fr> wrote:
> Bruce Griffiths <bruce.griffiths at ...> writes:
>> Are you sure that the setup is aligned to minimise AM response?
>> Are you using cross-correlation?
>> On Monday, 22 December 2014 5:08 PM, Loïc Moreau <loic.moreau <at>
> eai..fr> wrote:
>> Hi all,
>> My phase noise measurements system give erroneous results in close in
> phase noise measurements, I got
>> humps in the 1Hz-1000Hz area as high à 20 dB more than expected.
>> The setup is using a mixer to compare reference and DUT witch drive an LNA
>> http://www.ko4bb.com/~bruce/LowNoiseMixerPreamp.html , the output is
> driving an AD7760 ADC and an op
>> amp circuity is connected to the VFC reference to achieve quadrature. The
> results are analyzed with an
>> homebrew FFT charting software
>> After struggling with different configurations, switched different LNA,
> ADC , sound card. A scope
>> connected to the LNA output indicate steady 20MHz residuals just before
> the ADC ( around 10 mv peak-peak).
>> In fact , it seems that the mixer 20Mhz residuals ( DUT + REF ) are
> entering the ADC and so theses alias give
>> erroneous results in the 1Hz-100Hz area, displaying unexpected
> artifacts. In order to fix the problem
>> I will probably include an analog filter just before the ADC input (same
> as LNA input 1nF 80µH), but I
>> want to know if some more sophisticated measures should be undertaken as
> an 5th Order Lowpass Filter.
>> I have not found many clue about alias problems in phase noise
> measurements literature so I may have
>> missed something ?
>> Any advice
>> time-nuts mailing list -- time-nuts at ...
>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-
>> and follow the instructions there.
> I have a simple setup to measure the noise floor, the mixer is a SYM-2 and
> a divider is feeding RF and LO inputs ports for that purpose.
> a pi network give the 90° phase difference.
> The test is running with an HP 10811 giving around 7dBm, but i have the
> same results with any sources, especially a 33521A witch can be adjusted
> in level and frequency.
> I intend to use cross-correlation later when i will be confident with my
> the setup, for now i run a simple FFT.
> I have no clue about AM noise problem, i suppose that a sufficient input
> level on LO will put that problem aside.
> by the way, I was a bit surprised that nobody pay any attention to RF
> leakage from multiplier as the low frequency level is order or magnitude
> lower than 20Mhz product present at the input.
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
More information about the time-nuts