Signal Processing / EM / Radar Research Blog
L-band microwatt radar front end
May 2, 2008: Senior Design pre-kick off meeting
We discussed the following topics summarized below:
* Currently, the emphasis is on developing a Labview/NI-DAQ based solution that provides cancellation of interfering sinusoids, with the software controlling hardware to do so * The radar sends a periodic linear FM ramp waveform. The tones it receives due to reflections off objects in its view will be converted by Labview into a 1D display of distance. * The relationship between frequency and range to target (for each target) is as follows: Range_frequency = Radar_bandwidth / BW_ramp_time * round_trip_time <--> Fr = BW / Tm * Tr <--> Tr = Fr * Tm / BW and distance = 3x10^8 * Tr / 2 Example: Radar transmits ramp wavefrom from 2.0-3.0GHz, the ramp waveform having a period of 5 milliseconds. A spherical target exists, and the sinusoidal frequency received by the radar due to reflection off the sphere is observed at 37kHz. So, BW=1GHz, Tm=5ms, Fr=37kHz Thus, Tr=37k * 5m / 1G = 185 nanoseconds round-trip time. Knowing the speed of light is 3x10^8, we see that: distance = 3x10^8 * 185n/2 = 27.75m So, the sphere must be 27.75 meters from the radar antenna. * The radar leaks energy from its transmit antenna into its receive antenna that appears as an extremely strong low-frequency sinusoid at the NI-DAQ card. We intend to feed a Labview-controlled amplitude scaled and phase shifted version of the transmit waveform into the receiver, 180 degrees out of phase. This should mostly eliminate the "coupling" which is a significant hindrance to linear FM radars.
The above image is a low-res screen capture in the time-domain showing the very strong coupling. You see the blue time domain sinusoid, with the faint brown waveform showing it in the frequency domain. Note that the squiggles in the blue waveform are from the higher frequency radar returns (actual targets). * We'd also like to cancel strong radar returns coming off of closeby objects. This would involve another amplitude scaling and phase shifting, and a Labview controlled oscillator that is moved to the frequency of the offending object. We search to find the matching amplitude, frequency, and 180 degree phase shift to cancel the interferer. * Pulse radars shut the receiver off during transmit, but the problem we discussed is that since radar energy travels 30cm (1 foot) per nanosecond, it is a problem to generate pulses that narrow and switch that fast. Linear FM radar uses some physics tricks to enable high-resolution radar imaging at short distances (but linear FM becomes less useful at long range because range frequencies become too high to feasibly sample due to Nyquist rate). * A real system would likely want to use embedded processing, but we may not have enough time to implement it since the core focus is on the previous experiments.
Feb 2008: Signetics "Write Only Memory" advertisement
Advertisement for "Write Only Memory" published in April, 1974
February 2008: Michigan Nextel Cell Tower site locations
This is a low-fidelity representation of data I collected over the 2001-2005 timeframe of Nextel cell site locations. Each blue dot represents the location of a cell site. It took me a great deal of time to capture and collate this data.
Sept 2, 2007: Updated D-Star repeater co-channel discussion
At http://www.mvolz.info/dstar/#coch
Aug 31, 2007: Updated D-Star repeater discussion
At http://www.mvolz.info/dstar/#channel
July 20, 2007: HF Circuit reliability with VOACAP and NEC2 antenna modeling
Added http://www.mvolz.info/hfprop/, which describes an initial HF circuit analysis that is found to be unreliable.
June 22, 2007: Microwatt radar
After being shelved the better part of a year due to not having a functional digitizer, the microwatt L-band radar system was revived via use of a borrowed digitizer. I had used it fall 2006 with an oscilloscope at the microwatt level to see humans at over 50 ft in the CW mode, but now I was finally able to use it in the FMCW mode to get more serviceable data in terms of what might be used for scene imaging. And it works very well despite the highly cluttered indoor test environment.
Apr 23, 2007: Voltage References
A snippet of a conversation with an engineering student...
I would suggest running separate return (ground, negative) wires for each power supply, even though they tie in common at the PIC end. This will help control the amount the ground voltage pulls away from 0 Volts under varying loads. This is because of the finite resistance of wires and soldered connections that becomes a measurable real-world quantity with heavier currents, especially at the low voltages you're dealing with.
For my radar system where I am very sensitive to any disturbances or fluctuations, is where despite all boards having more or less a common mechanically established ground reference, for important signal paths I ran separate grounds too. I also put a power regulator on each board which gives a firm relationship between V+ and ground locally for that board.
I am careful to avoid power flowing through a signal ground conductor, since that changes the voltage away from 0V. Well, not so careful, I don't have the boards mechanically isolated, but I try to provide low impedance paths for power to help keep it out of my signal grounds. It is upsetting to me to see systems where a mechanical or signaling ground is used purposefully to send power through. If you're lucky, it will work initially but as oxidation sets in, flaky things can happen.
Jan 27, 2007: Discussion of expectations for radio coverage predictions
A snippet of a conversation I had....
I took a look with the -110dBm to the mobile at 50% location and 10% time, at desired/undesired ratios of 17dB and 10dB.
So, then the question comes down to what defines a covered area, and what defines interference. Also, one must consider the need to ensure that a full communications loop could exist. That standard may not be met with a -110dBm talkout standard.
No one standard is right or wrong for every situation, it's what the user expectations are that's important. Naturally amateur radio users are more forgiving than in commercial or government systems.
Where did -102dBm come from? A figure based on extensive feedback from end-users and empirical testing that allows enough margin for multipath, interference, etc. Considering amateur radio needs and expectations, this figure might be altered downward...or maybe not.
What are we actually fighting against besides anthropogenic interference that is a physically constant quantity? Johnson-Nyquist or thermal noise on the channel. We might consider an FM receiver of 16kHz bandwidth, a typical FM amateur radio (e.g., ± 8kHz from channel center). We might reckon this receiver considering its own noise figure has a noise floor of -126dBm. This comes from realizing the thermal noise floor in dBm as:
-198.6dBm/Hz-K [Boltzmann's constant]+10·1og10(300K) [ambient temperature]+10·log10(16000Hz) [receiver bandwidth]+6dB [radio NF]=-126dBm
So from here we might say, why isn't -109dBm or -110dBm enough since that's 17dB over the noise floor? --it can be, if you're in the outback of Austrailia operating off a standalone car battery with no other electronics around for some miles. But most sites/locations have more than thermal noise present, especially at VHF. We also want to be sure to have enough signal to overcome cable and duplexer losses, which while lowering environmental noise, also push us closer to the immutable thermal noise floor. Thus -110dBm could be too low a signal level in real-world situations.
Interference between stations takes multiple forms:
1) Interference of one repeater transmitter to other repeater's base/mobile stations
2) Interference of one repeater's base/mobile stations to other repeater's receiver.
Typical standards used in the Longley-Rice predictions for the desired coverage area for 5W ERP omnidirectional 20ft. high base stations was defined as a median -102dBm signal from the base TO the desired repeater at 50% of the locations in a cell 90% of the time.
Typical standards used in the Longley-Rice predictions for the desired coverage area for 25W ERP mobiles at an antenna height of about 5ft was defined as a median -102dBm signal from the mobile TO the desired repeater at 90% of the locations in a cell 90% of the time.
Interference for mobiles and base stations was defined as a desired/undesired ratio of less than 17dB as is well established for 5kHz system deviation FM systems; 50% of locations 10% of the time.
Jan 25, 2007: Modes of variability for Longley-Rice predictions
With the release of Radio Mobile Deluxe, propagation modeling is now within the reach of individual scholars. There is a downside to this availability, in that results may vary wildly without basic knowledge of key parameters. It takes a great deal of time and some study to understand the necessary principles.
Terminology varies and the inner working of these modes are generally not well understood--here is a brief listing of the four modes of variability:
Types of Variability:
Location: Treats discrete cases where the intervening terrain profile data is the same, e.g. two locations 50 meters apart, each 10 kilometers from the base station.
Time: Any long-term variations in signal strength--specifically EXCLUDES variations due to multipath etc. Caused by changes in refraction index etc.
Situations: Accounts for cases where you have the same location and equipment, but for natural reasons otherwise unaccounted for, the signal level varies.
Modes of Variability:
Broadcast: Treats all three kinds of variability individually, useful for where a "hard look" is wanted.
Individual or Accidental: Combines location and situation variability, considering those two combined and time variability--useful for focusing on time variability from a single location.
Mobile: Combines location and time variability, since logically for a mobile a change in location translates to a change in time. Considers location/time combination and situation variability--e.g. driving the same path repeatedly you would expect such-and-such confidence.
Single-message or Spot: This collapses all variabilities into a simpler model that considers what the chance of a single attempt succeeding is.
Useful for mobile-to-mobile predictions.
References
Hufford, G., Longley, A., Kissick, W. "A Guide to the Use of the ITS Irregular Terrain Model in the Area Prediction Mode." 1982. http://www.its.bldrdoc.gov/pub/ntia-rpt/82-100/
Longley, G. "Radio Propagation in Urban Areas." 1978. http://www.its.bldrdoc.gov/pub/ot/ot-78-144/
Jan 7, 2007: Coverage area prediction discussion
After a long hiatus improving the L-band system and preparing to present academic papers, I've got a new snippet for the blog.
This entry concerns standards of two-way radio coverage
Having a frequency offset is generally a good technique to increase spectrum reuse, however, typically one must be 10kHz or more away to start seeing a significant benefit, due to the bandwidth of the receivers and also the width of the FM modulation. 5kHz is generally not enough to provide substantial benefit with 25kHz channel spacing FM receivers.
Oct 3, 2006: VHF waveguide
Well, I looked at using a 55 gallon barrel for an open-ended circular waveguide antenna on VHF. Sorry, diameter is too small--you don't get above lowest mode cutoff until around 300MHz.
Sept 30, 2006: Poster presentation
The poster I presented to the Dean and faculty on Sept. 30 is archived on the Radar Projects Page.
Aug 4, 2006: Powerpoint presentation
The Aug 4 presentation was a success.
Radar Projects PageJuly 27, 2006: Updated Block Diagram
In light of the continued success of the design, we decided to change the implementation to an even more modular design to enable on-the-fly switching of components as needed for high resolution applications.
New Block DiagramJuly 18, 2006: Progress continues
Hardwiring S-band radar in preparation for usage on sampling rail. Acquired NI DAQ card and Labview software.
July 16, 2006: Updated PC-Radio Interface
Added audio diagrams and further explanation.
July 15, 2006: NASA Shuttle audio rebroadcast system diagram
At: http://www.mvolz.info/em/w8sh/july06/w8msu_audio.html
July 14, 2006: VHF Interference study
Upon commissioning a link on the 2 meter (VHF) amateur band, I conducted a study to ensure that harmful interference to a repeater 70 miles away would not occur. Results at http://www.mvolz.info/rf/coverage/
July 13, 2006: Radar is working
Radar is working, albeit without image processing hooked up yet. Using no receive amplificaton, I was easily able to pick up reflections at 20 feet, as viewed on a noisy old oscilloscope. This is actually very promising, as it seems to be very sensitive.
July 10, 2006: Audio bandwidth considerations
New page at: http://www.mvolz.info/radio/diag/. Radar construction is going very well, just no time to update here!
June 20, 2006: Board error
Oops, my board design left the output connectors too close, it would have been impossible to screw the plugs onto the jacks. My personal licence of Sonnet didn't allow for a sufficient matrix size to predict a new, slightly larger layout, so off to the Sun Blade 1500 workstation with the full Sonnet licence....
June 19, 2006: Board damage analysis
During construction, one of the boards was accidentally damaged on the groundplane side, quite near the microstrip line. I was concerned over this, but empirical tests evinced no frank ill effects from the damage.
I then conducted a quick simulation of the defect, and from a quick examination, no frank effects relevant to the system appear to be predicted.
Charge Density
June 19, 2006: Power Divider Design
Designed a new splitter based on information from various texts, mainly Gupta's. It is wonderful to have the software design resources here at State--the simulations came out ideally.
Current Density
S-Parameters--Note that S21 and S31 overlap entirely on this plot
June 16, 2006: Wilkinson Power Divider
My original plan of using a surface mount toroidal splitter was dashed when the one unit I had obtained was defective. Rather than deal with another wait to get another, and for the challenge, I decided to construct a Wilkenson power divider.
I had first used a set of approximation functions from Bahl's A Designer's Guide to Microstrip Line, Microwaves, May 1977. They worked fine for my initial microstrip design, but for a confidence check I ran another set of piece-wise approximations I obtained from Gupta's Microstrip Lines and Slotlines.
These approximations [note: large image--must view at full size for clear text] show the results of the piece-wise approximation vs. the first set of approximations I used. The results are clearly adequately consistent and provide additional confidence in the predictions.
June 14, 2006: Photo of modules
These are the modules I have built thus far...more testing and finalizing in the days ahead...
June 12, 2006: Correspondance
I had this email correspondance with a chap who asked about interference on low-orbiting FM satellites:
...The figure of 17dB minimum C/N for acceptable S/N quality comes from empirical tests based on what end users would find an “acceptable” carrier-to-noise for analog FM modulation. It usually works out to be about 30dB S/N with 25kHz bandwidth FM radios (at a desired/undesired ratio of 17dB). 20dB S/N is where with no modulation, the crackle (but not hiss) on the received signal just disappears. 12dB SINAD has noticeable crackling. (Of course, to measure SINAD, one must use a proper SINAD meter). For 30dB S/N, the channel is "full quieting" under steady conditions. This 17dB figure changes signifcantly between modulation types such as digital or SSB, AM.
To see if you have enough Carrier/Noise (CN0) (i.e. >=17), you use this formula.
CN0 = (PC - AG) – (RS – CNS) ;
Where:
PC = average power in dBm of desired signal with NO interfering signals
AG = Antenna gain + feedline loss
RS = 12dB SINAD sensitivity level of receiver (in dBm)
CNS = static carrier-to-noise ratio of receiver, typically 4 or 5 for modern FM 25kHz receivers.
This merely relates the receiver sensitivity to the actual signal level. Subjective empirical tests are done to ascertain this level, especially true for digital audio such as HD Radio etc.
Now, you may be interested in free space loss (also consider Faraday rotation and ionospheric losses which are NOT in this equation!). A simplified version of Friis’ equation—which is only accurate in the Fraunhofer (far-field region). The Fraunhofer region is defined as R>(2·D2)/λ; where D is largest antenna dimension (in meters) (give λ in meters too).
Simplified version of Friis:
Path loss = 20·log10[(4·π·d)/λ] ; where d is distance between antennas in meters, and λ is in meters.
You will soon find there is around 9dB more loss on 440MHz than on 146MHz, and this brings some interesting considerations to mind when designing satellites for portable users. I believe that having 146MHz uplinks is bad in the sense that the lower path loss means omnidirectional interlopers can much more easily wipe out the uplinks to the satellites. But, 440MHz uplinks would be a little harder for the end user to know how to manually tune for Doppler. But I think having no interlopers would more than make up for that difficulty.