Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711

E-mail: Jack.Smith@cliftonlaboratories.com
 

This page documents some very quick phase noise measurements I made of five signal sources. The sources are:

• HP 8640B analog signal generator
• Panasonic VP8191 digital synthesized signal generator
• Z90 in signal generation mode
• HP87510A VNA in CW mode
• HP8752B VNA in CW mode

The measurements were made at 13.5 MHz, with an Advantek R3463 spectrum analyzer and a home made 13.5 MHz four-crystal notch filter. The idea is that if we notch the carrier down 70 dB or so, we can measure the residue with a spectrum analyzer without worrying so much about the spectrum analyzer's internal phase noise characteristics.

Updated 09 June 2007. I have included phase noise measurements made when visiting W4XN's lab. The same process was used, except for an updated crystal notch filter. The updated notch design is at 8 MHz (nominal) and has about 4 dB insertion loss. Mike's data was captured with an HP8566A analyzer, my new design 8 MHz notch filter, and the following signal generators:

• HP 8640B analog signal generator
• HP 654A analog function generator
• HP 3325A synthesized function generator
• HP 8660C synthesized RF signal generator with each of the following plug-in units:
  • HP 86601A (0.01 to 110 MHz)
  • HP 86602B (1 to 1300 MHz)
  • HP 86603A (1 to 2600 MHz)

This is the Z90 output at 13.5 MHz. The undesired signal seen are approximately 65 to 70 dB (the two strongest spurious signals) below the desired 13.5 MHz output, which is notched down 75 dB or so. This sort of noise is characteristic of a DDS, and can be seen in the AD9851 data sheet.
 

The VP8191 is a digital synthesized signal generator covering from 100 KHz to 135 MHz, with built-in AM and FM modulation, output settable in 0.1 dBm steps and frequency settable in 100 Hz increments.

The data shows two discrete spurious signals about ±1.6 MHz from the carrier and a substantial noise pedestal about 1 MHz wide. The discrete spurs are down about 90 dB or more from the carrier, after the 70 dB notch is subtracted. The notched carrier is clearly seen about 10 dB above the noise pedestal. This shape is characteristic of a PLL system.
 

The HP8640B is widely known for its low-noise characteristics. These derive from the instrument's design, which runs a cavity-tuned oscillator in the 256-512 MHz range, with lower frequencies derived by digital frequency dividers, followed by octave low pass filters.

Sure enough, the HP8640B comes through. The noise level is about 115 dB below the carrier level  and the spurious signals are approximately -95 dB below the un-notched carrier. Compare the 8640B with the Panasonic VP8191.
 

The 87510A is a Vector Network Analyzer (officially called a "gain/phase test set, but it's really a VNA) covering the range 100 KHz to 300 MHz. It also can be used as a signal source by setting the sweep width to 0 Hz.

It generates a series of reference spurious sideband, at about 600 KHz intervals and rather high noise level.

HP8752B in CW Mode

The HP8752B is a Vector Network Analyzer covering the range 300 KHz to 3 GHz. As with the 87510A, it can be operated as a signal source by setting the span to 0 Hz.

There is a fair bit of broadband noise and also two strong spurs at ±1250 KHz. The spurs are about 60 dB below the un-notched carrier level of 0 dBm. It is superior to the 87510A in both spurious and broadband noise level.
 

The 8640B's output is +15 dBm, and the noise level is less than -90 dBm. You can just see a carrier pip where the crystal filter was unable to completely notch it out. Still, the notch depth is around 105 dB, which is not too shabby, indeed.

As measurements on my 8640B and Mike's 8640B show, these are superb generators with respect to phase noise. Mike's has broadband and phase noise at least 105 dB below the generator output, as that is the floor of this particular measurement setup.

HP 654A is HP's wide ranging RC-tuned oscillator design of the 1960's, covering from 10 Hz to 10 MHz.

As the data below shows, it produces much more more phase noise than the gold standard 8640B.  The 654A was set to -15 dBm output, 30 dB less than the 8640B.