A Segmented Chirped-Pulse Fourier Transform Millimeter Wave

Spectrometer (260-290 GHz) with Real-time Signal Averaging

Capability

Brent J. Harris, Amanda L. Steber, and Brooks H.PateDepartment of Chemistry

University of Virginia

Segmented CP-FT: Single Spectrum Acquired in 100 ms

100 µs, 35 GHz (260 – 295 GHz)

Simple Benchtop Prototype

65 cm sample cell

Advantages for Chirped Pulse mm-wave Spectroscopy

A. Steber, B. Harris, J. Neill, B. Pate, J. Mol. Spec 280 (2012) 3-10

1) Rapid spectrum acquisition (35 GHz spectrum acquired in 2 ms)

2) Coherent measurement methods for spectrum analysis

3) Transition frequency accuracy

4) Simple spectrum signal processing (no baseline effects)

Challenges for Chirped Pulse mm-wave Spectroscopy

1) Real-time signal averaging to achieve high sensitivity

2) Reduction of spurious signals

3) Spectral resolution and Fourier transform signal processing

Instrument Design Principles: The Chirped Pulse

1) The Chirped Pulsed Separates the Pulse Duration from the Pulse Bandwidth

High-Resolution Spectroscopy: Line width is much smaller than frequency range

Line width: set by the signal dephasing (Doppler, Collisional Relaxation)

Optimal Excitation Pulse Duration: ~ 1 / (line width)Transform Limited Pulse Duration: ~ 1 / (frequency range)

2) In the Weak Pulse Limit the Signal is proportional to (Bandwidth)-1/2

Signals for transform limited pulses scale with (Bandwidth)-1 :

Spectral energy density reduced by both increased bandwidth and shorter pulse duration.

Instrument Design Principles: Segmenting

Segmenting the Measurement Frequency Range

Relative Segment Bandwidth: (1/N)

Rel .Signal in Segment: (2)½ Reduction in Averages: ½ Rel. Measurement Time: 2 * ½ = 1

Full Bandwidth CP-FT

Rel. Signal in Segment: (N)½ Reduction in Averages: (1/N)Rel. Measurement Time: N * (1/N) = 1

Things that are segment independent (all things being equal):

1) The time required to reach a target dynamic range (signal-to-noise ratio)

2) The number of points that make up the spectrum

For smaller segments, lower digitizer rates can be used

(Transform limited pulses: increasing the segment bandwidth increases the measurement time!)

Instrument Design Principles: Segmenting

Segmenting the Measurement Frequency Range

Things that are segment dependent (all things being equal):

1) The time required for one spectrum scan (N times longer)

2) The total number of data points processed (Reduced by N)

Molecules perform some of the averaging

Things that are not equal:

1) Digitizer cost and real-time signal averaging performance (favors more segments)

Still need bandwidth to use power efficiently and minimizing stitching

2) Proliferation of spurious signals (favors more segments)

Exception: Single Segment has best mm-wave practical spur performance

260-295 GHz CP-FT Design for Real Time Averaging

Number of Spectrum Points: 35 GHz frequency range, ~ 2ms dephasing timeMinimum Points: (70 Gs/s) x (2 ms) = 140,000

This number of data points can be accommodated in FPGA accumulator512K points maximum, 8-bit digitizer, 32-bit data size

Maximum real-time accumulations: 2(32-8) = 224 = 16M

Minimizing the Spurious Signals: Strongly Dependent on AWG purity for LO generation

LO Impurity Spurs: LO-to-IF Conversion (self mixing): Easily subtracted

Spectrum Images: Improved AWG performance on N*30 MHz

This fixes the segment bandwidth to 30MHz x 24 = 720 MHz: 50 segments, 100 ms scan time

Digitizer/Receiver Spurs: Second Order IM: work in 2nd Nyquist zone (720 – 1440 MHz) Third Order (Two-tone IM) Signal/Clock Mixing in Digitizer

Segmented Chirped-Pulse Fourier Transform Spectroscopy

Separate AWG Channels Generate Chirp Segments (Blue)

and

Local Oscillator (LO) Frequency (Red) with Phase Reproducibility

Output Power: 30-40 mW

2-3.5 GHz

10.8-12.3 GHz 260– 295 GHz

220 – 325 GHzWR 3.4

Low IF (720 – 1440 MHz)4GSDigitizer

High Speed Segmented CP-FT Measurement

Digitized at an IF of 720 – 1440 MHz at 4GS/s

Real-Time signal averaging in 32 bit FPGA :524288 sample memory (131 µs, at 4GS/s)

2 µs segments:- 0.250 µs chirped pulse- 1.75 µs decay time- 720 MHz bandwidth (per segment)- 50 segments (36 GHz, 100 µs )

Spectrum Relative Intensity Performance

Frequency dependent spectrometer response:- Source power- Receiver mixer loss- IF amplifier gain- Propagation loss

System response correction:- Routine response curve

generated with series of stored single frequency waveforms

- Correction incorporated into signal processing

About 10-15% accuracy

Intensity Reproducibility

Intensity variation < 0.1%

Suitable for monitoring time evolution of the broadband spectrum

Comparison to Literature mm-wave BenchmarkS. Fortman, I. Medvedev, C. Neese, F. De Lucia, ApJ, 725 (2010) 1682CH3CH2CN 0.5 mTorr, 6 m path length, 24 s equivalent measurement time

Speed: CP-FT is 10,000 times fasterIntensity Accuracy: 15% vs. 1% (but 0.1% precision)Line width: 2.5 broader in CP-FT (magnitude FT + windowing)

High Dynamic Range Mode (HDR)

HDR sequence:

- 0.250 µs chirped pulse- 1.75 µs decay time- 24 MHz bandwidth/seg- 30 segments for each LO

frequency (720mHz)- 31st segment for electrical

background subtraction- 50 LO frequencies (1500

total segments)

62 ms for 720 MHz

Two 720 MHz segments can be accommodated in the FPGA

Noise Floor Achieved in 1s: 1 mV (60,000:1 dynamic range)

Broadband Double Resonance Spectroscopy

Acrolein

MeasurementProtocol:

1) Single Frequency Pump Pulse

2) 720 MHz Chirp

Observe ~60% signal modulation

Sequential and V-Type Level Schemes Show Different Modulation Behavior

Conclusions

1) Real-time Signal Averaging is Achieved using Segmented Chirped Pulse Fourier Transform Spectroscopy

2) High Speed Mode Significantly Reduces Measurement Times Over Absorption Spectroscopy Methods

3) A High Dynamic Range Mode Provides Spurious Signal Reduction and Achieves about 100,000:1 (Largely) Spur Free Performance

4) Capabilities for Broadband Double Resonance Spectroscopy Can Translate Reduced Spectrum Acquisition Times to Reduced Spectrum Analysis Times

Acknowledgments

Brent Harris is supported by an NSF Graduate FellowshipUniversity of Virginia Equipment Trust FundNSF I-Corps