Using the HP8970A with circulators.
(March 9 2013)

The purpose of using circulators with NF meters.

It is well known that the impedance of a noise head changes between the on and the off state. That causes the gain of the LNA under test to change and that change affects the Y-factor (the power ratio) that the instrument computes the NF from.

The error caused by the impedance variations of the noise head can be minimized by making the input impedance of the amplifier under test very close to 50 ohms.

For development of a front end for use on 1296 MHz it is desireable to be able to compare NF values with very high accuracy because of the extremely low sky temperature on this band. Only about 3 K. To clearly see the effect of changes it is desireable to be able to measure noise temperatures within 0.1 K which corresponds to 0.0014 dB on the NF.

That kind of accuracy is not obtainable even by usage of matched amplifiers. The best amplifiers use GaAs FETs. Those devices are voltage sensors that absorb very little energy. Without feedback the input return loss would be close to zero. Nearly all the energy from the antenna would be reflected back to the antenna by the LNA.

By use of noiseless feedback one can feed a suitable amount of the amplified signal back to the antenna to cancel the wave reflected back by the FET. The price to pay is that the gain becomes lower which makes the noise contribution from the second stage larger.

The setup with circulators.

By use of two cascaded circulators one can easily isolate the noise head by 80 dB by some tweaking with magnets. That provides adequate isolation from the impedance variations and allows the usage of a noise head with 15 dB ENR. The circulators have to be followed by an attenuator to isolate the test object from the circulators. Without the attenuator the test object would otherwise see extremely high SWR at large offsets from the frequency of operation which could lead to oscillations.

The attenuator for which a suitable value is 3 dB has to be thermally isolated from the test object by a cable of suitable length. Otherwise the temperature of the attenuator will be affected when different test objects are connected to it. By use of a semi-rigid cable one can also use this cable to tune the impedance at the test port to 50 ohms precisely. Figure 1 shows the set-up.


Figure 1. The circulator arrangement for precise NF measurements.


The base plate is 20 mm thick aluminium, four layers of 5 mm aluminium bolted together. The noise head is in good thermal contact with the base plate. three thermally conducting compressible isolators intended for TO3 transistors make sure that the aluminium body of the noise head is in good thermal contact with the base plate. A precision thermometer is attached to the front surface of the noise head. The platinum sensor is inserted with some silicon grease into a copper tube that is glued to the noise head. The 3 dB attenuator has another precision thermometer. The semirigid cable from the attenuator has a variable capacitor that can be adjusted to squeeze the cable by the appropriate amount on the appropriate position along the cable.

Manual measurement.

With the below procedure one obtains noise figure readings that are reproducible witin about 0.02 dB provided that the temperature is nearly constant. The procedure requires that the ENR value of the noise head does not change with temperature (Which is not true below about 23 degrees C for the particular noise head I have studied) and that the attenuation of the circulators, the attenuator and the cables does not change with the temperature. These assumptions are reasonable in the temperature range 23 to 28 degrees with the particular setup I have been using.

The set-up for the procedure is like this:

  • 1. Calibrate with the noise head directly on the input of the 8970A. Make sure the noise head temperature is stable and feed the temperature into the instrument before calibrating.


  • 2. Verify ENR: Enter 15.2, then "special function" and check that the ENR value is correct. Change if required.


  • 3. Loss before the DUT: Enter 34.2, then "special function" and check that the loss before DUT is correct for your setup. With the setup of figure 1 I have been using 4.6 dB for the total losses of the circulators and the 3 dB attenuator.


  • 4. Loss after the DUT: Enter 34.4, then feed in the attenuation of the cable used to connect the DUT to the 8970A.


  • 5. Set loss compensation ON: Enter 34.1, then "special function"


  • 6. Set arithmetic smoothing: Enter 13.1, then "special function"


  • 7. Set display to Te: Enter 10.4, then "special function"


  • 8. Set temp input to Centigrade: Enter 11.1, then "special function"


Use the buttons on the 8970 to set the desired number of averages and to set the frequency of operation. Then feed in the correct temperatures for the noise head and for the attenuator after the noise head. by use of the special functions 6.0 and 34.3 With an averaging number of 512 the reading should be stable to within 0.1 K provided that the temperatures are stable enough to not change by more than 0.1 degree over the time span of several measurements.