Sec 5 Signal processing - TSI

pancakesbootΤεχνίτη Νοημοσύνη και Ρομποτική

24 Νοε 2013 (πριν από 3 χρόνια και 11 μήνες)

86 εμφανίσεις

TSI Incorporated

Copyright© 2008 TSI Incorporated


Signal Processing




TSI LDV/PDPA Workshop & Training


Presented by Joseph Shakal Ph.D.

TSI Incorporated

Copyright© 2008 TSI Incorporated

Outline


Nature of the Signal


Processor Requirements


FSA Architecture
-

Front End & Burst Detector


FSA Architecture
-

Samplers


FSA Architecture
-

Firmware Processor


Burst Centering


Dynamic Sampling Rate Selection


Other Features


Conclusion

TSI Incorporated

Copyright© 2008 TSI Incorporated

Nature of Doppler Signals



Amplitude not constant



Lasts for only a short time, which itself varies



Amplitude varies from burst to burst



Presence of noise



High frequency



Random arrival

TSI Incorporated

Copyright© 2008 TSI Incorporated

Signal Processor
-

Key Requirements


Use multi
-
bit sampling up to a high maximum frequency


Detect and validate bursts based on SNR and amplitude


Automatically optimize the sampling rate for each burst


Gives the best resolution in processing, even for a wide range of velocities


This will ensure the maximum number of cycles are used


Detect the burst center, before processing


Use the data from the middle portion of the burst first


Use this value as the arrival time of the particle


Detect burst duration separately


Digitize and record additional analog and digital signals, including
cyclic markers and the burst amplitude

TSI Incorporated

Copyright© 2008 TSI Incorporated

FSA Signal Processor


Uses 8 bit sampling up to a 800MHz


Detects and validates bursts based on patented real
-
time SNR
measurement, and also amplitude


Automatically selects the optimum sampling rate for each burst,
which is also a patented technique


Detects the burst center, since processing is done after sampling
process is complete


The FSA EB option digitizes and records additional signals,
including cyclic markers


Burst amplitude is measured as part of the patented Intensity
Validation technique

TSI Incorporated

Copyright© 2008 TSI Incorporated

Nature of Input Signal

Out of PMT Detector

Out of PDM

Noisy Signal

TSI Incorporated

Copyright© 2008 TSI Incorporated

Impact of Type of Burst Detection

Signal

Method of Burst Detection


Source

Amplitude

Amplitude/

Envelope Based

SNR Based

Small Particles

Small

Ignores

Detects

Large Particles

Large

Detects

Detects

Surface
Reflections

Large

Detects

Ignores


Other Noise,
Spikes, etc.

Any

Detects some

Ignores


TSI Incorporated

Copyright© 2008 TSI Incorporated

The FSA Signal Processor

TSI Incorporated

Copyright© 2008 TSI Incorporated

FSA Processor

Burst Detector Subsection

LUT/DFT

Amplitude

Threshold

Controller

Signal Out

Burst
Gate

Downmixer

Downmix
Frequency

Ch 1

Frequency
Estimate

Dynamic Optimum
Sampling Rate
Selection

Burst Gate

Filters

TSI Incorporated

Copyright© 2008 TSI Incorporated

FSA Processor

Burst

Sampling Subsection

Controller

Multibit

A/D

Sample

Memory

Signal Out

Burst
Gate

Downmix
Frequency

Frequency
Estimate

Burst Gate

Optimally
Sampled

Data

Firmware

Processing

LUT/DFT

Ch 1

Filters

TSI Incorporated

Copyright© 2008 TSI Incorporated

FSA Processor

Firmware Processing

DSPs

Burst
Processing

Firewire

Interface

Optimally
Sampled Data

Burst gate &

Burst center

to PC

TSI Incorporated

Copyright© 2008 TSI Incorporated

FSA Block Diagram Summary

to PC

Signals from PDM

Downmixer

Downmix

frequency

generator

Amplitude

threshold

Controller

8 bit

A/D s

Sample Memory

DSPs

Signal Out

Burst Gate

Firewire

Interface

Bandpass

filters

Ch 1

Ch 2

Ch 3

Frequency
estimate

Burst
gate

EIC/EB

pressure, temp

OPR or Shaft Encoder

LUT/DFT

TSI Incorporated

Copyright© 2008 TSI Incorporated

Burst Detection and Sampling


Burst Detector


Determines the approximate burst frequency


Determines the beginning, end, and center of the burst


Burst Sampler


Dynamically selects the optimum sampling frequency
(using the approximate frequency value) so that each and
every burst is sampled at the optimum rate


Obtains the 8 bit digital values, and passes them on to the
Firmware Processing subsection

TSI Incorporated

Copyright© 2008 TSI Incorporated

Burst Centering

Burst centering is automatically done by the FSA, and it helps give higher
quality data by using only the high
-
SNR portion of the signal for processing.

The center point of the burst is identified.

From this reference point, samples are used out to a certain noise threshold, until
the FSA’s data block is filled.

Gate Time (transit time) is still based on the actual time the particle was in the
measurement region, regardless of the portion of the signal used for processing.

Beginning

End

Center of

burst

This portion used for
processing

TSI Incorporated

Copyright© 2008 TSI Incorporated

Dynamic Sampling Rate Selection

Example

Burst
gate

Particle 1


velocity = u


sampling rate: F

Particle 2


velocity = 2u


sampling rate: 2F

Particle 3


velocity = 4u


sampling rate: 4F

Burst gate

Burst gate

TSI Incorporated

Copyright© 2008 TSI Incorporated

Flow and Size Analyzer (FSA)

Other unique features and benefits of the FSA


Built in input buffer, for high data rate situations


Size measurement validation


Patented intensity validation uses an independent measured quantity
to validate the diameters. We do not need receiver masks, and we
can see what is being rejected and what is being accepted.


Phase validation, uses the degree of agreement between the two
independent phase measurements to discriminate between reflection
and refraction


Short transit time flows (50ns minimum gate time)


Particle size measurements in dense sprays, where we need very
small measuring volumes, resulting in short
-
transit
-
time burst signals


Size measurements in (pulsed) high velocity sprays


Velocity measurements in supersonic flows

TSI Incorporated

Copyright© 2008 TSI Incorporated

Measured Parameters





Ch. 1

Velocity Mean (m/sec)

20.390

Velocity RMS (m/sec)

1.4284

Turbulence Intensity (%)

7.01

Frequency Mean (MHz)

6.3207

Frequency RMS (MHz)

0.4428

Frequency TI (%)


7.01

Gate Time Mean (usec)

2.26

Gate Time RMS (usec)

1.32

Data Rate (Hz)


43245

Valid Count


2663

Invalid Count


0

Elapsed Time (sec)


0.0616

Statistics

Plots

TSI Incorporated

Copyright© 2008 TSI Incorporated

Conclusions


Examined the nature of Doppler signals


Looked at processor requirements


FSA Architecture
-

Front End & Burst Detector


FSA Architecture
-

Samplers


FSA Architecture
-

Firmware Processor


Saw how burst centering works and its benefits


Looked at how dynamic sampling rate selection is done and its
benefits


Other benefits of the FSA, like high
-
speed capabilities for dense
sprays and high speed flows, and FireWire connectivity