[SOLVED] [Moved]: Phase noise increases with reduction in oscillation frequency

Hi,

I fabricated an IC that has a quadrature VCO and I am measuring the phase noise. I observed in simulations that when I change the varactor voltage to change the frequency, the phase noise reduces (i.e. phase noise performance improves) as I reduced the oscillation frequency. Why does this generally happen?

When I measure the IC for the phase noise performance, I observe two things.

1) The phase noise is way worse than I what I simulated. I am supposed to get around -126 dBc/Hz @ 1MHz for a 5GHz oscillation frequency. This number is supposed to reduce further, as I reduced the oscillation frequency to let's say 4GHz by tuning the varactor voltage. What I observe is that I get around -113 dBc/Hz@1MHz for a 5GHz signal and around -107dBc/Hz@1MHz for a 4GHz signal. I have attached the images. What could be the possible reason for the increase in the phase noise number with reduction in oscillation frequency?

Re: Phase noise increases with reduction in oscillation frequency

venn_ng said:

The phase noise is way worse than I what I simulated.

Click to expand...

Phase noise is almost impossible to accurately simulate. I would suspect that your increase in phase noise as you drop in freq is due to divider noise.

Re: Phase noise increases with reduction in oscillation frequency

SLK001 said:

Phase noise is almost impossible to accurately simulate. I would suspect that your increase in phase noise as you drop in freq is due to divider noise.

Click to expand...

Hi,

Thanks. The thing is this is a free running Quadrature VCO so I don't have a PLL.

Re: Phase noise increases with reduction in oscillation frequency

venn_ng said:

Hi,

Thanks. The thing is this is a free running Quadrature VCO so I don't have a PLL.

Click to expand...

You VCO is probably drifting by Temperature.Have you ever looked at the carrier at zero span SA with low sweep time ??
Because PN curves are not normal..

Re: Phase noise increases with reduction in oscillation frequency

Hi,

Thanks for the reply. I will check with zero span and let you know. What's not normal with the PN curve? You mean that it goes flat close to 10kHz?

Re: Phase noise increases with reduction in oscillation frequency

BigBoss said:

You VCO is probably drifting by Temperature.Have you ever looked at the carrier at zero span SA with low sweep time ??
Because PN curves are not normal..

Click to expand...

Hi,

I obtained some plots with zero span and different RBWs. The total signal power is around -20 dBm. The signal is spread across a bandwidth of 100kHz based on the plots. I have included some images below. Please let me know why you feel the phase noise plot is not normal.

Re: Phase noise increases with reduction in oscillation frequency

venn_ng said:

Hi,

I obtained some plots with zero span and different RBWs. The total signal power is around -20 dBm. The signal is spread across a bandwidth of 100kHz based on the plots. I have included some images below. Please let me know why you feel the phase noise plot is not normal.

Click to expand...

These figures don't show zero span.The x-axis will be the time not frequency in zero span measurements.Either do zero span or decrease RBW to zero if applicable..
As you see, the output power is not constant, so how the PN will be measured under this circumstance ? While you're decreasing RBW, the spread becomes obvious.
Phase Noise is a relative measurement as you know well and the measurement equipment compares the Relative Power @ Offset to Carrier Power.So since the carrier power is not constant, how this measurement will be consistent ??
Every measurment will be different then previous one because the carrier power is shaking..I still think the VCO is drifting and even the center frequency is catched once, second time it will be shifted slightly.

Re: Phase noise increases with reduction in oscillation frequency

BigBoss said:

These figures don't show zero span.The x-axis will be the time not frequency in zero span measurements.Either do zero span or decrease RBW to zero if applicable..
As you see, the output power is not constant, so how the PN will be measured under this circumstance ? While you're decreasing RBW, the spread becomes obvious.
Phase Noise is a relative measurement as you know well and the measurement equipment compares the Relative Power @ Offset to Carrier Power.So since the carrier power is not constant, how this measurement will be consistent ??
Every measurment will be different then previous one because the carrier power is shaking..I still think the VCO is drifting and even the center frequency is catched once, second time it will be shifted slightly.

Click to expand...

Hi,

Thanks for the reply. I don't understand why this is not zero span and the x-axis is also time and not frequency. As far as the phase noise measurement is concerned, I am interested in phase noise at 1MHz. So the phase noise at 1MHz offset doesn't change with measurement time, which is what I observed. In addition, the analyzer updates the carrier after every decade of phase noise measurement I guess based on what I saw during measurement.

Re: Phase noise increases with reduction in oscillation frequency

venn_ng said:

Hi,

Thanks for the reply. I don't understand why this is not zero span and the x-axis is also time and not frequency. As far as the phase noise measurement is concerned, I am interested in phase noise at 1MHz. So the phase noise at 1MHz offset doesn't change with measurement time, which is what I observed. In addition, the analyzer updates the carrier after every decade of phase noise measurement I guess based on what I saw during measurement.

Click to expand...

OK, let's assume your arguments are correct. The frequency is not shifting...
But what about the power ?? Are you sure the Power @ Offset Frequency is following this roll ? I cannot be sure..
You have changed the RBW and you found 4 different result that are quite different from each other.Does it make sense ??
If the signal quality was good, even the RBW is changed, the power should be-approx- same, shouldn't it be ??

It may be a power supply problem, don't underestimate that.If there is a remarkable noise on the supply line, it disturbs too..

Re: Phase noise increases with reduction in oscillation frequency

The analyzer measure the total carrier power correctly because it calculates over the entire bandwidth.
The power at the offset frequency is not going to be constant if the carrier is changing. Based on zero span measurements, we can see that the carrier power remains constant if the RBW is 1MHz but changes a lot if the RBW is 100kHz. So I assumed that phase noise @1MHz offset should be close to its real value. Based on measurement results, I found that even though the phase noise close to 100KHz is changing every time(I made continuous sweep), the phase noise at 1MHz offset doesn't and remain almost the same during each measurement. So I believe the phase noise @ 1MHz offset is accurate.

Re: Phase noise increases with reduction in oscillation frequency

BigBoss said:

OK, let's assume your arguments are correct. The frequency is not shifting...
But what about the power ?? Are you sure the Power @ Offset Frequency is following this roll ? I cannot be sure..
You have changed the RBW and you found 4 different result that are quite different from each other.Does it make sense ??
If the signal quality was good, even the RBW is changed, the power should be-approx- same, shouldn't it be ??

It may be a power supply problem, don't underestimate that.If there is a remarkable noise on the supply line, it disturbs too..

Click to expand...

Hi,
Regardless of the phase noise, there is a huge variation in the center frequency (100 kHz) with time. What do you think can generally cause this issue? Supply noise or I am using an external LDO, can that also cause that noise? Is there a way to measure phase noise accurately with that variation?

Re: Phase noise increases with reduction in oscillation frequency

The PN measurement method employed by your spectrum analyzer won't give you accurate results for the VCO with frequency drift. Search for NIST/HP technical notes about PN measurement. You will need to use a reference VCO with known/better PN and lock it with low-bw PLL to your VCO. Then by analyzing voltage from the phase detector you can estimate PN of your VCO.

Hi.

Looks like my LDO (off-chip) was very noisy and that was causing some issue with the phase noise measurement due to frequency drift. Now that got fixed. Thanks for your help.

venn_ng said:

Hi.

Looks like my LDO (off-chip) was very noisy and that was causing some issue with the phase noise measurement due to frequency drift. Now that got fixed. Thanks for your help.

Click to expand...

You're right, voltage which supplies the VCO must be very clean, has to be noise-free as much as possible..
But there is another problem on your circuit.Even the LDO is noisy, the VCO shouldn't be disturbed too much as much as 100 kHz.The pushing figure seems too bad.
Check also decouplling capacitors' quality and surely GND connection around the VCO.I think there are still some second degree issues.

Read about "aperture jitter" as the time-domain counterpart
of phase noise.

When you decrease frequency (of a sine wave) you also
decrease the zero-crossing slew rate and increase the
time spent within the linear aperture of the phase
detector's logic. This increases the time during which
voltage noise is converted to "time noise" across the
input slope (jitter or PN, depending on the instrument
used).

If the aperture becomes wide enough that output
activity can wrap back to the phase detector, then
things could get really icky.