Sub-Millimeter RF Receiver balanced using Polarization Vectors Intrel Service Company iscmail@intrel.com Sub-Millimeter Week of RF 19Receiver August 2012 Copyright Intrel Service Company 2012 Some Rights Company, Reserved Slide 1 c Intrel Service
What is Sub-Millimeter RF? Sub-mm wavelength uses Radio astronomy Short-range/space communications material identification c Intrel Service Company, Slide 2
Terahertz Semiconductor Receivers What do terahertz receivers look like? c Intrel Service Company, Slide 3
They use FETs/Diodes & Waveguides Small, complex, & delicate Waveguides: 145 X 310um, 3um tolerance or 5.7 X 12.1mils,.12mil tolerance c Intrel Service Company, Slide 4
They need Cryogenic Cooling Example SIS Mixer (Superconductor-Isolator-Superconductor) 290 K 77 K 12 K 4 K c Intrel Service Company, Slide 5
Sensitivities are far from "Photon Noise Limit" 780-920 Ghz atmospheric window Noise temperature Receiver: 250 K Photon noise limit: 24 K Receiver operating temperature: 4 K c Intrel Service Company, Slide 6
A Better Solution: µwave/optics-hybrid Thermal Photo-Mixer Use microwave technology to get the Local Oscillator (LO) Use a mixer for gain Balance mixer to null LO noise Thermal detector from optics technology Operate at room temperature (!) Use a large, 3x3mm detector to match f/5 diffraction spot Eliminate: tiny waveguides and components dipole antenna and coupling network critical dimensions and distances c Intrel Service Company, Slide 7
What is photo-mixing? How does it work? Is there a suitable detector for these wavelengths? c Intrel Service Company, Slide 8
Photo-Mixer Concept It is just interference fringes! Bright signal Dark nothing Beams aligned: good Put the detector here (Airy's disk from circular aperture) Beams tipped: bad (Must align to better than 2.8 degrees) c Intrel Service Company, Slide 9
Mixer Math Review Antenna E1*sin( 1*t) + E2*sin( 2*t) Power Detector (pwr~e2) E12*sin2( 1*t) + 2*E1*E2*sin( 1*t)*sin( 2*t) + E22*sin2( 2*t) Math Handbook E1*E2*[cos( ( 1+ 2)*t ) - cos( ( 1-2)*t )] Mixer Gain (pwr1*pwr2) c Intrel Service Company, Slide 10
Detectors for 300µm and 300 K Type Golay cell Thermopile Bolometer Pyroelectric Opto-Mechanical* Principle gas expansion Seebeck effect resistance charge storage solid expansion D-Star f3db(hz) 1.6e+9 3 2.0e+8 10 2.0e+8 200 3.0e+8 750 3.5e+16 900 *The OMD uses a Mechanical Expansion Amplifier (US Patent 7,707,896) c Intrel Service Company, Slide 11
Q: What intermediate frequency (IF) will the mixer generate? A: The IF frequency is at zero Hertz, sometimes called homodyne conversion rather than heterodyne conversion. Q: The frequencies below the LO appear at negative frequencies. What happens to them? A: They fold over to positive frequencies, superposing the lower sideband onto the upper. This works fine for molecular radiation and most modulation techniques used in communications. c Intrel Service Company, Slide 12
Homodyne (Direct) Detection Difference: where the IF filter is placed Thermal detector bandwidths 1.0 to 100KHz Radio astronomy, secure communications, signatures c Intrel Service Company, Slide 13
Q: Isn't oscillator phase noise a problem operating so close to the LO frequency? A: Yes, indeed it is. c Intrel Service Company, Slide 14
1.0THz Local Oscillator: Oewaves OEO 10 GHz spectrum. Multiplied to 1000 GHz. c Intrel Service Company, Slide 15
Q: What is mixer gain and what does it do to LO noise? A: The next slides illustrate the effects of mixer gain in 3 steps. Q: What is the goal? A: Show that mixer gain makes the proposed detector as sensitive as a state-of-the-art detector. Represent the sensitivity level of the state-of-the-art detector by a signal at the corresponding power level, _Sig. Show that mixer gain can bring that signal above the inherent detector noise, MEA_. Q: What does the first slide show? A: It shows the relative power levels of the detector noises (MEA_, Instr_, LO_n_) and the signal to be unburied (_Sig). The mixer gain is one, no gain. c Intrel Service Company, Slide 16
Start: Mixer with Low LO Power OMD detector, NEP ~ 2.5e-12 LO noise -40dbc (Watts) Start with Local Oscillator at target, P_lo= P_sig= 1.5e-18W/ Hz Mixer gain = 1.0: sqrt(p_sig*p_lo)/p_sig c Intrel Service Company, Slide 17
Q: What happens when the LO power is increased enough to amplify _Sig to MEA_? A: The blue arrow on the right shows the increase of LO power. The middle blue area shows the buried signal increasing amplitude in proportion to the square root of the LO power. On the left, the detector noises are unaffected. But the LO noise (LO_n_) increases along with the LO. Q: Isn't it bad news to have the LO noise greater than the signal and detector noise? A: Sure is. Something has to be done about that. c Intrel Service Company, Slide 18
Homodyne with LO Add LO Increase LO until Sig = OMD Noise Note LO noise. Oops! Need LO noise suppression c Intrel Service Company, Slide 19
Q: What are the red downward arrows in the next slide? A: They show the decrease in LO and LO noise power when a balanced mixer is used. The change represents a balancing to one part in 10,000. The balanced mixer configuration also increases the signal power by a factor of 2. Q: Is balanced mixing effective for semiconductor diode, FET, or bolometer detectors? A: Not very. It is difficult to get the two detector gains matched to much better than 5%, creating subtraction to only one part in 20. Q: So something special is done to get the 10,000 to 1 subtraction? A: Yes, indeed. c Intrel Service Company, Slide 20
Homodyne with Balanced LO Balancing brings noise below mixer (OMD) noise Preferred method: calc LO noise after balancing increase LO until LO noise = mixer noise use Sig-to-OMD margin to expand bandwidth c Intrel Service Company, Slide 21
Q: Balanced mixers are commonly used at radio frequencies. What is so different for sub-millimeter frequencies? A: The ±90 degree phase shifts required for balancing can be obtained using passive circuit components at low frequencies. At terahertz frequencies the phase shift is obtained from the path length difference to the two detectors. The path length and thus the phase shift is not very stable. Q: What does a state-of-the-art radiant wave balanced mixer look like? A: The next slide show the configuration used with lasers or microwave beams. c Intrel Service Company, Slide 22
Traditional Balanced Photo-Mixer Changes in any pink path-length will change the mixing phase. keep length << wavelength beam angles must be almost perfect, much less than 2.8 c Intrel Service Company, Slide 23
Q: Is there any solution to the path-length stability problem in terahertz balanced mixers? A: Yes, a novel polarization method was invented, designed, built, and tested by James A. Kuzdrall in 1968. Q: How is path-length phase change prevented? A: The LO and signal are combined on a mixing half-mirror before splitting to travel to their separate detectors. The relative phases are locked to each other at the same phase difference in both beams. There can be no path length induced phase changes to create noise. Q: If the relative phases to the two detectors are locked, how is the ±90 degree phase obtained? A: Using polarization, a radiant wave parameter which is completely independent of temporal phase, but can be used in the same way. c Intrel Service Company, Slide 24
Polarization Photo-Mixer Layout Signal & LO joined at Half-Mirror Relative phases locked! Signal converges to Airy's disk Any telescope f/# allowed No signal waste c Intrel Service Company, Slide 25
Polarization Vector Arithmetic Polarization resolves into a parallel and a perpendicular component on a polarizer, in the usual vector way. The parallel projection passes through while the perpendicular component is absorbed or reflected. Polarization vectors in instantaneously different directions subtract. In the same direction, they add. The LO is linearly polarized in a fixed direction. The polarizer axis are at +45 and -45 degrees to the LO polarization direction. The signal is randomly polarized, resolving into equal components along and perpendicular to each polarizer. The signal polarization changes over time, sometimes in the direction of the LO projection and sometimes in the opposite direction. There are 4 possible combinations of LO and signal polarization directions. c Intrel Service Company, Slide 26
Local Oscillator, all Cases The red lines in Cases 1 through 4 which follow represent the local oscillator. It always has the same amplitude and orientation. Equal amplitude projections (+1.5) appear on the left and right polarizers. They both go in the same direction, up. When the outputs of the right and left detectors are subtracted, their local oscillator responses will exactly cancel, along with their noise. The local oscillator output cancels for all 4 cases, as desired. The oscillator noise cancellation being accomplished, can it be shown that the signals do not also cancel? In Case 1, they do. Not a good start. c Intrel Service Company, Slide 27
Signal Vector Results, 4 Cases The green lines in the Cases represent the signal to be detected. Random polarization, the most common, will at any instant resolve into components parallel and perpendicular to the polarizer axis. At that instant each resolved component may be going in either of 2 directions. Four possible combinations of resolution angle and direction must be explored, creating the 4 Cases. In the first two Cases, the signal outputs from the two detectors cancel when subtracted. In the last two Cases, the signal outputs add, giving a signal output. The outputs of Cases 3 and 4 also add, giving a total response equal to the input signal. Just what was desired! c Intrel Service Company, Slide 28
Case 1 of 4, No Output Signal components that resolve in-line with LO give no output (Signal is assumed to be randomly polarized.) c Intrel Service Company, Slide 29
Case 2 of 4, No Output Signal component that resolves in-line with LO gives no output (Signals are assumed to be randomly polarized) c Intrel Service Company, Slide 30
Case 3 of 4, Signal Output Transverse polarization resolves to an output Signal peak amplitude is 2/2 before being "resolved" c Intrel Service Company, Slide 31
Case 4 of 4, Signal Output Transverse polarization resolves to an output Signal peak amplitude is 2/2 before being "resolved" c Intrel Service Company, Slide 32
Q: Why do Cases 3 and 4 indicate a non-zero result rather than a number? A: The vectors represent the electric field intensity on the OMD detector. These must be converted to heat by squaring. The the mixer gain must also be taken into account. Q: Why is there confidence that Cases 1 and 2 are zero? A: Both the signal and LO signals have the same amplitude and direction. Whatever the results, they will be equal, subtracting to zero. For Cases 3 and 4, the signals on the right and left polarizers go in opposite directions. The results cannot be equal. Q: If the math were worked out, what would the subtracted outputs in Cases 3 and 4 be? A: Twice as much as that of a single mixer. c Intrel Service Company, Slide 33
A Practical Polarization Mixer Phase Locking Plate Half-Mirror Detector 1 18KHz Beam Chopper (Transmitter) Collection Lens Detector 2 Laser (NeHe) Coherent transmitter/receiver using polarization balanced mixer (~1968, Sanders Associates) c Intrel Service Company, Slide 34
Balanced Mixer Performance Return signal modulation (~18KHz) Both mixers (balanced) Laser power supply ripple (~20KHz) Bottom trace, single detector: One mixer (unbalanced) peak on left is ripple on laser (LO) output from 20KHz power supply peak on right is the 18KHz signal Top trace, balanced detector: LO ripple component is canceled 18KHZ signal is ~2x amplitude c Intrel Service Company, Slide 35
Balancing the Detectors Q: Can LO noise be used to balance the detectors? A: The broadband rms noise does go to its lowest value at balance, but the lowest value is complex for instrumentation to judge autonomously. Q: What else is there to use? A: Nothing, so something must be added. In this scheme, the LO is modulated with a sine wave located near the upper bandpass limit. After subtraction, the phase of the sine wave output depends on which detector has the highest gain. A synchronous detector produces an output that passes through zero. It drives a potentiometer which nulls both the sine wave and the LO noise. Q: Won't the modulation sine wave mask legitimate signals? A: The sine wave is always nulled to zero. c Intrel Service Company, Slide 36
Auto-Balancing Circuit Typically 1000 to 20,000 suppression c Intrel Service Company, Slide 37
SMRR Receiver Performance c Intrel Service Company, Slide 38
Thermal Homodyne Summary Simpler hardware and electronics 300 K operation Covers all wavelengths, 100um to 1000um 300GHz to 3THz Practical balancing for LO noise suppression Superb sensitivity c Intrel Service Company, Slide 39
(These antennas could really use our help!) Contact: James A. Kuzdrall iscmail@intrel.com c Intrel Service Company, Slide 40
References www.intrel.com/mea/index.html "Laser Receivers", Monte Ross, Wiley 1966, 66-13513 "Radar Handbook" Merrill Skolnik, McGraw-Hill 1990, ISBN 0-07-057913-X c Intrel Service Company, Slide 41