Need expert's help on Frequency Synthesizer Spur

Hi guys,

I am measuring a fractional-N PLL based frequency synthsizer these days, and I encounter a tough spur problem as below:

the synthesizer is used in a wireless SoC for RF applications, it is found that there are two -20dBc spur (located@+/-700kHz offset) at the synthesizer output, it is quite large;
when the SAR ADC(Successive Approximation Register ADC) is OFF, the 2 tones disappears, or when the SDM(Sigma Delta Modulator) is OFF, the 2 tones disappears too; the SAR ADC clock is divided down from VCO frequency(which is to say the spur is not due to the intermodulation between harmonic of ADC clock and VCO);
the SAR ADC share the same supply with synthesizer(with RC filtering on each supply side) except VCO, the SDM is in the digital baseband side;
the location of the 2 tone depends on the fractional divide ratio, the smaller fractional ratio, the nearer offset is;
no matter how I increase the XO harmonic power(by heavily increasing the buffer size of XO buffer), the spur doesn't change;
it is not fractional spur;(99% percent)
it is not because of magnetic coupling;(because we have both LC VCO and ring VCO version, both have the 2 spur)

anyone can give me some clue? thanks very much!

Best Regards
Buckaroo

buckaroo said:

Hi guys,

I am measuring rf frequency synthsizer these days, and I encounter a tough spur problem as below:

the synthesizer is used in a wireless SoC for RF applications, it is found that there are two -20dBc spur (located@+/-700kHz offset) at the synthesizer output, it quite large;
when the SAR ADC is OFF, the 2 tones disappears, or when the SDM is OFF, the 2 tones disappears too; the SAR ADC clock is divided down from VCO frequency(which is to say the spur is not due to the intermodulation between harmonic of ADC clock and VCO);
the SAR ADC share the same supply with synthesizer(with RC filtering on each supply side) except VCO, the SDM is in the digital baseband side;
the location of the 2 tone depends on the fractional divide ratio, the smaller fractional ratio, the nearer offset is;
no matter how I increase the XO harmonic power(by heavily increasing the buffer size of XO buffer), the spur doesn't change;
it is not fractional spur;(99% percent)
it is not because of magnetic coupling;(because we have both LC VCO and ring VCO version, both have the 2 spur)

anyone can give me some clue? thanks very much!

Best Regards
Buckaroo

Click to expand...

Hmmm ... just two spurs +- 700 kHz ... visual indication from the spectrum analyzer (?) would indicate an AM-type Modulation (amplitude modulation) of the signal, as opposed to FM modulation (think: Bessel Functions and multiple carriers) .. so I would conclude in a first-order analysis that the 'coupling' or 'method of ingress' or effect of the unwanted signal is via the Vdd (or Vcc) power supply line, maybe to a buffer stage? If the signal generation is digital in nature, there would seem to be am amplitude function being applied at some point, unintentionally.

I'm not 100% clear on what kind of synthesizer you are working with (PLL or DDS) or what SAR (successive approximation register?) and SDM mean, so my above first-order analysis may not bear much weight.

In any event, as a check, can you look down at the milliVolt level with a short ground lead scope probe (ground right at the scope probe tip) at the ripple on the DC supply lines under your two test conditions?

Jim

RF_Jim said:

Hmmm ... just two spurs +- 700 kHz ... visual indication from the spectrum analyzer (?) would indicate an AM-type Modulation (amplitude modulation) of the signal, as opposed to FM modulation (think: Bessel Functions and multiple carriers) .. so I would conclude in a first-order analysis that the 'coupling' or 'method of ingress' or effect of the unwanted signal is via the Vdd (or Vcc) power supply line, maybe to a buffer stage? If the signal generation is digital in nature, there would seem to be am amplitude function being applied at some point, unintentionally.

I'm not 100% clear on what kind of synthesizer you are working with (PLL or DDS) or what SAR (successive approximation register?) and SDM mean, so my above first-order analysis may not bear much weight.

In any event, as a check, can you look down at the milliVolt level with a short ground lead scope probe (ground right at the scope probe tip) at the ripple on the DC supply lines under your two test conditions?

Jim

Click to expand...

Hi Jim,

Thanks for your reply.
If the coupling is from the power supply(on-chip LDO), it seems hard to generate -20dBc tones(-20dBc is large), accurately I have measured the power supply frequency spectrum from the spectrum analyzer, I can only see large spur @high order harmonics of SAR ADC clock(=divided down from VCO frequency fvco) and VCO frequency, the spur @fvco+/-700kHz cannot be seen;(the vco frequency fvco is about 1.6GHz, I directly pulled out the power supply from on-chip LDO and connected to spectrum analyzer after ac coupling.)
it is a PLL-based frequency synthesizer, SDM means the Sigma Delta Modulator to control the divider to get a fractional divide ratio; SAR ADC is Successive Approximation Register ADC, which use fvco/N as its clock in my case;
any suggestions? thanks

Rds
buckaroo