Noise Parameter Basics. Dr. Zacharia Ouardirhi Dipl. Ing. Matthias Beer MBA

Transcription

1 Noise Parameter Basics Dr. Zacharia Ouardirhi Dipl. Ing. Matthias Beer MBA

2 Presentation Outline Noise Figure vs Noise Parameter Noise Parameter Extraction Noise Parameter Measurement Setups Noise Parameter Verification 2

3 Noise Figure Definition G a S N in in + G a S N out out N a G a +N a Output Signal = Input Signal * Gain: S.out = S.in * G Output Noise = Input Noise * Gain + Internal Noise: N.out = N.in * G + N.dut Noise Figure is the Signal/Noise Ratio Degradation through the media (DUT) NF SNR SNR Input Output 3

4 Noise Figure vs Noise Parameters To design a Low Noise Amplifier you need the four Noise Parameters, not only the Noise Figure Noise Circles Stability Circles Gain Circles This transistor has a noise figure of 2.9dB at 50 but a minimum noise figure of 0.85dB at 100+j150 4

5 Noise Parameter Measurement Setup Noise Source F.total Input Tuner The tuners synthesize.source F.rec Receiver LNA Noise Receiver 50.source F.dut = F.total (F.rec 1)/Gav.dut A two port device has 4 Noise Parameters To determine them at least 4 x.source are needed Focus method uses many more (~50) 5

6 Extract DUT Noise Figure Using FRIIS formula for a cascade of amplifiers with Noise Figure Fi and Available Gain Gi We derive F.total = F1 + (F2 1)/G1 + (F3 1)/(G1*G2) F.dut( s)=f.total( s) {F.rec( dut( s)) 1}/G.dut( s) where: dut( s) = S22 + (S12*S21* s)/(1 S11* s) s DUT {Sij} Receiver F.rec dut = f( s) 6

7 Four Noise Parameters Noise Figure depends on Ys as follows: F (Ys) = Fmin + Rn/Gs * Ys Yopt 2 This is the equation of a parabolic hyperboloid surface with a set of isometric circles on the Ys Smith Chart, with F as parameter 7

8 Determination of 4 NP 8

9 Wideband Stability s Stability Circle (: s for dut=1) ATTENTION The DUT may oscillate at low frequencies, even though we measure at high frequencies, because the tuner is wideband! opt > Fmin Tuning Range Min Freq Max Freq 9

10 Wideband Impedance Spread s Stability Circle (: s for dut=1) For fast noise measurements, the tuner has to move least, this means the impedance points cannot be optimum at all frequencies NFmin Tuning Range Min Freq Max Freq 10

11 Source Impedance Patterns Low Density Pattern 1 (12 21 points) = spiral starting at S11* at fmin and fmax and fmed if fmax/fmin> 4 High Density Pattern 2 (25 50 points) Same as above S11* S11* Target Pattern 3 (25 50 points) = triangle centered around S11*, else as above Customized tuning pattern 11

12 Selecting Valid Noise Data Monitor various (user defined) criteria to maintain the quality of the measured data. Maximum s Filters out any measured point with s > Limit Mismatch limit Mismatch Factor shows the distance of s from the input stability circle of the DUT. The larger the number, the closer s to the stability circle. Oscillation Avoidance Monitors the fluctuations in DC current to detect instabilities. Skip the current measurement source impedance at all frequencies. 12

13 Selecting Valid Noise Data Monitor various (user defined) criteria to maintain the quality of the measured data. Maximum s Filters out any measured point with s > Limit Mismatch limit Mismatch Factor shows the distance of s from the input stability circle of the DUT. The larger the number, the closer s to the stability circle. Oscillation Avoidance Monitors the fluctuations in DC current to detect instabilities. Skip the current measurement source impedance at all frequencies. 13

14 Noise Data Processing use SVD Singular Value Decomposition (SVD) In linear algebra, the singular value decomposition (SVD) is an important factorization of a rectangular real or complex matrix, with many applications in signal processing and statistics. Applications which employ the SVD include computing the pseudo inverse, least square fitting of data, matrix approximations, and determine the rank, range and null space of a matrix Use at least 8 measured impedance points 14

15 Effect of Filtering on Noise Data Raw Data Filter Filter&Smooth 15

16 Noise Measurement Setup 1 VNA is measuring S Parameters of DUT and NFAnalyzer the NF VNA NFA Source Tuner ENR INM DUT ONM Focus Focus INM: ONM: Input Noise Module Output Noise Module 16

17 Noise Measurement Setup 2 VNA is measuring S Parameters and NF of DUT VNA Source Tuner INM Focus DUT ONM Focus INM: ONM: Input Noise Module Output Noise Module 17

18 Noise Parameter Measurement Set Up with ZVA only* * above 24GHz additional accessories for ZVA option K30 are required Tuner INM Focus DUT ONM Focus INM: ONM: Input Noise Module Output Noise Module Control LAN Focus 18

19 Noise Parameter Measurement Set Up with PNA X (Opt.029) +28V J11 J10 J9 J8 J7 rear panel J2 J1 + Source 2 (optional) OUT 1 OUT 2 Noise receivers R1 Source 1 OUT 1 OUT 2 Pulse modulator Pulse modulator 10 MHz 3 GHz GHz A LO R2 B To receivers Test port 1 Source 2 Output 1 Source 2 Output 2 Test port 2 Tuner DUT Noise receivers 19

20 Noise Measurement Setup (PNA X) The PNA X has two noise receiver options: 1.Option 029: True Noise Receiver (based on Noise Source and 8975) 2.Option 028: RMS/AVG (signal only) Focus supports both methods Wideband tuner GHz to match the PNA X 26 20

21 Broadband Noise Parameter System Components Control Module LAN/TCPIP Output Module Input Module 21

22 Load Pull and Noise System Low Frequency Noise Parameter System GHz 22

23 Load Pull and Noise System High Frequency Noise Parameter System 10 67GHz 23

24 Ultra Wideband Noise Parameters GHz 24

25 Noise Parameter Verification 25

26 The problem A load pull test setup can be verified using the back to back (B2B) method: one tuner conjugate matches the other one at many points using a THRU as a DUT. If the Gain is 1 then the system is accurate. * 26

27 Load Pull Setup Verification Gt=0.03dB 27

28 In Noise things are not so easy Noise Parameters are not measured, they are calculated Noise model: F = Fmin + Rn/Gs* Ys Yopt 2 4 accurate noise measurements would be enough to determine Fmin, Rn, Yopt=Gopt+jBopt This is never the case, measurements in general and noise measurements in particular are noisy Therefore we use > 4 points and process the data Assuming the hardware is correct, the whole noise measurement literature over the last 50 years has been about Choosing the right Ys points Properly processing the data 28

29 What is a Noise Standard A Standard is a device of which we know, ahead of time, the exact physical properties like the 4 noise parameters A practical noise standard is one users can produce themselves. Passive networks, without noise sources other than thermal, are such standards. Data sheets measured by other systems are not noise standards, they are simple references for comparison. 29

30 Why are passive networks Noise Standards? It comes from the definition of Noise Figure: Signal Ys {S ij } Signal/Loss Thermal Noise Thermal Noise Noise Figure F = SNR IN / SNR OUT F (Ys) = Available Loss = 1/Gav(S ij,ys) Calculate F for an arbitrary set of {Ysi} 4NP 30

31 Accuracy of Noise Parameter F(Ys, S ij ) =1/G av =( 1 s*s 11 2 *(1 o 2 )/( S 21 2 *(1 s 2 )) where o = S 22 + S 12 *S 21 /(1 s*s 11 ) s = (1 Ys/Yo)/(1+Ys/Yo); Yo=20mS. 31

32 Test Equipment used: Receiver 1: Miteq LNA1 (up to 18GHz) + PNA X Receiver 2: Miteq LNA2 (up to 26GHz) + PNA X Receiver 3: Miteq LNA2 + R&S Spectrum analyzer. 32

33 Noise Data Processing Theory Raw Data (3 samples) Statistical Processing, Fmin Statistical Processing, Rn 33

34 Matched Attenuator 3dB Enable Post Extraction Optimization 34

35 Attenuator 6dB 35

36 Attenuator 9dB 36

37 Why is that important? The raw data (measurement) have been correct The true result is contained in the raw data Statistical processing used to extract the final data has been correct There cannot be a coincidental agreement over such a wide frequency range. 37

38 Mismatched passive Standard 38

39 S parameter of mismatched Standard High pass Filter Limited accuracy 39

40 N Par of mismatched Standard Theory 40

41 Noise Parameters to verify F50 41

42 N Par Verification thru F50 Measuring Fdut at 50 = F50 (Tuner initialized) Calculating F50 from extracted 4 N Par F50 = Fmin + Rn/Gso * Yso Yopt 2 Yso=Gso+jBso: Tuner Initialized 42

43 Verification thru F50 Measuring F Dut at 50 = F50 ( tuner init) Calculating F50 from extracted N Par F50.c F50.m Yso(f ) F50.c F50.m +0.45dB 0.45dB F50 = Fmin + Rn/Re(Yso) * Yso Yopt 2 Initialized with Tuner 43

44 Conclusions Focus offers Wideband noise parameter solution The widest band tuners Support of VNA, Spectrum Analyzers and NFM Passive noise standard testing allows Verifying the raw data Verifying the Extraction Verifying the Statistical Math (Averaging) F50 testing allows Consolidating active device data 44

45 Focus Microwaves Winning through Innovation