My RF/DSP projects and publications

July 25, 2016
    I have created this page to share my knowledge related to electronics and computer programming. Materials provided here might be useful for students, engineers and other researchers. My intent is to organize my knowledge, get some feedback, show my capabilities to potential employers.

For any questions and/or comments, please email me at

CURRENT PROJECT: Advanced microwave sensor
Main goals: cost effectiveness, DFM, advanced features allowing multiple applications (hobbyist, automation, security, etc.)


December 2, 2016: Using STM32 to acquire FSK signals.
    Proof of concept firmware was developed to acquire FSK signals using STM32F407 microcontroller. Further research will be performed later.

November 21, 2016: Proof of concept doppler sensor prototypes testing
    Proof of concept doppler sensor prototypes are being tested. Acceptable performance is achieved. The only issue is analog power supply noise, which will be filtered out in next prototype. VCO layout need to be adjusted to achieve better S21 phase matching.
Doppler sensor prototype

November 11, 2016: STM32F4 preliminary firmware framework is ready.
    Buffered ADC / DMA code is ready, it is specially optimized to perform modulation and signal acquisition of different waveforms: pulsed, frequency shift keying (FSK), frequency modulated continuous wave (FMCW) and multiple frequency continuous wave (MFCW). Only FSK modulation is planned in current project because of VCO limitations. Next step is to perform some tests on PCB with all crucial elements soldered on: RF part and DSP part. STM32F4 Discovery board was used to develop preliminary firmware framework, which could be readily be used on a custom designed board.

STM32F4 Discovery

USART of STM32F4 Discovery is connected to PC through ADM3251E IC which provides full isolation for RS232 port.

November 10, 2016: Moving code from STM32F103 to STM32F407.
    Added most important parts from my previous project to the new one: USART code, performance measurement code, FLASH routines, FFT routines. Need to move ADC code and write new DMA code. Compared results from STM32F407 and PC FFT code to ensure accuracy. With help of STM32CubeMX and HAL libraries work is done really fast.

November 7, 2016: Moving from Turbo Delphi to QT
    Today I have ported simple but important piece of code I used previously in Delphi. It is FFT code for performing two real-valued FFT transforms using single complex FFT transform. QTCustomPlot was used for spectrum visualisation (

Testing FFT transform in QT Creator

    I also have checked simple webcamera code which I already made before using QWidget promoted to QCameraViewFinder. The only part left is to learn some simple RS232 communication. QTCreator is a fascinating editor, with this piece of software development becomes much easier. Next step is testing preliminary prototype and porting some old code from STM32F103 to STM32F407.

November 5, 2016: Moving from Turbo Delphi to QT
    I have used Turbo Delphi programming tool for many years: making tools for interacting with different devices over serial port, programming different DSP algorithms (adsp218x microcontrollers), and also 3D graphics algorithms as a hobby. A pretty time consuming part was porting DSP source code written in Pascal language to assembler or C language. To improve my productivity I decided to move from using Delphi to QT. It will allow to copy code between platforms in a much easier way. Now I can write C-style algorithms on a PC platform and easily reuse them as source code for ARM microcontrollers.

November 2, 2016: DSP board preliminary design
    Some parts of DSP circuit board are already designed. In my previous project i have used STM32F103RBT6 microcontroller. Although it had enough computing power for my task, firmware design was too complicated because of many optimizations and lookup table algorithms. In this project I decided to use STM32F407VET6 microcontroller, which is much more powerful with a little additional cost.
     In current project STM32F407VET6 microcontroller would be used to
1) Provide frequency modulation signal for VCO using DAC output (PA4:DAC_OUT1)
2) Acquire IF output from mixers using ADC inputs (PC4:ADC1_IN14, PC5:ADC1_IN15)
3) Perform signal processing using algorithms written in C language
4) Data exchange with host using UART port (PA10:USART1_RX, PA9:USART1_TX)
I configured microcontroller PLL to achieve maximum performance using 8 MHz HSE clock, HCLK is set to 168 MHz:

STM32F407 CubeMX clock configuration STM32F407 CubeMX pins configuration

Configuration is done using STM32CubeMX software.

October 28, 2016: Resistive mixer biasing notes.
    It seems that gate does not need to be biased very precisely. I obtained similar results for Vgs≈-1.3 v to -0.8 v.  Much more impact is from drain resistor value. Very useful information on active FET mixer design could be found in this appnote:

October 27, 2016: Research continued.
    After a little vacation I continued my research. Mixer IF output amplifier with automatic gain control was successfully tested. I checked improvised FET mixer performance using different biasing again. Further conclusions on mixer performance can be made only after prototypes are built, because current configuration lacks of high frequency decoupling. I am planning to use single-layer pcb for RF board prototype, and will use only one high-frequency signal transition VIA to minimize possible losses. VCO prototype must be built and performance of main RF parts evaluated during a tuning experiment. VCO control circuit must be designed with special requirement of signal acquisition synchronized with modulation impulses.

October 12, 2016: Dielectric resonator oscillator design.
    Currently I am designing an oscillator. Parallel feedback configuration was chosen, because it allows to use stable amplifier topology. Dielectric resonator (DR) will be coupled to drain and gate lines of a source-grounded FET transistor. To ensure that unwanted frequencies are damped, gate line will be terminated with 50 Ohm resistor. Output drain line will be matched to 50 Ohm output. DR will be coupled to open-circuited stub on a drain line, to provide better coupling between line and DR at the quarterwave distance from open-end of a stub. In previous project I placed DR between two 50-ohm terminated lines, which resulted in poor coupling and small output power:
Dielectric resonator oscillator (parallel feedback)
Gate stub was terminated with 50 Ohm resistor. In this design high impedance quartarwave open-circuited stubs can be added near resonator to improve coupling.
    The other two choices was series feedback oscillator and MMIC oscillator. Main problem with series feedback oscillator is high variation of FR4 dielectric constant value. Unwanted frequency of oscillation will occur in case if negative resistance achieved at frequency other than frequency of DR. In case of push-push topology this problem can be solved using less known parallel feedback push-push configuration, where two parallel feedback oscillators are combined. MMIC oscillators currently are too expensive, although could be considered for high-end designs or mass production designs. Cheapest MMIC with integrated mixer is approximately 100-times more expensive than single high quality high-gain FET transistor. Overall price of Dielectric Resonator and 3 FET transistors is approximately 6% of MMIC's price. Although MMIC oscillators have great advantage of frequency stability, wide band frequency modulation and meeting DFM requirement.
    In current design three main goals must be achieved: good frequency stability, good power output, low requirements to DR positioning. Frequency tuning element will be added later.

October 7, 2016: Design updates 2.
    Proof of concept resistive mixer was constructed using old diode mixer with hybrid coupler. Unused microstrip traces was cut, FET transistor was soldered instead of Schottky diode. To ensure that mixing is provided by FET, all Schottky diodes was totally soldered out. By quick tuning I found that gate biasing around Vgs≈-1.3v gives satisfactory performance. It is interesting to note, that IF noise reduced by 3dB, and conversion loss is not as high as expected. There was unwanted IF frequency of few kHz, it seems that it is some noise from power circuit. Amplitude of this noise varied periodically while tuning Vgs. The idea of resistive mixer at X-band proved to be feasible by experimental test. Very good performance was obtained, even without low-pass filtering on biasing line and without matching FET for X-band.

October 7, 2016: Design updates.
    Compared two designs of series patch array. First design uses 50 Ohm feeding line with deep inset (butterfly-shaped patches). Second design uses ~120 Ohm feeding line without inset. It is interesting to note, that using different feeding line impedance and varying inset distance allows change distance between patches without adding additional phasing delay lines (microstrip "curls"). Currently i prefer first design, as it gives lower side lobes level and more compact. Prototyping must be done to be sure if there are some potential problems, mainly with possible mutual coupling between butterfly-shaped patches.
    Mixer is planned to be single-ended FET mixer, or passive/active FET mixer. Both types will be prototyped. In case of single-ended FET mixer LO leakage problem must be taken care of. Passive/active FET mixer must give around ~ 30dB isolation, but require to use different signal path and separate TX/RX antennas. I am planning to use single FET device for oscillator and mixer, because it meets DFM goals. Using single FET device also reduces BOM components count, eliminating Schottky diode.
    Oscillator is planned to be DRO with FET transistor. Main disadvantage of using dielectric resonator is that it is not very DFM-friendly, as it requires positioning, gluing and tuning.  In future designs DRO may be replaced by other type of oscillator. Main reason of using DRO in current design is poor dielectric constant temperature stability of generic FR4 substrate.

October 6, 2016: Designing X-band series fed patch array.
    Two days was spent designing series fed patch array. Single patch was designed first. Scattering coefficients S11, S21 was analyzed to estimate amount of radiated power. Varying this amount can be used for weighting purposes to reduce side lobes or provide sharper beam. Patches are fed with 50 Ohm microstrip line. Unlike traditional 2x1 patch arrays used in cheap motion sensor, such array may provide higher gain and lower losses.
Single patch antenna element for series fed arraySeries fed patch array antenna
Usually very thin microstrip line is used to match high impedance on the radiating edge. On a FR4 substrate such approach seems to induce more radiation losses. Using large inset distance also allows to design several prototype to tune center frequency of array and using wide microstrip feeding lines. Small tilt in radiation pattern of array is exists but it is acceptable in most cases. Array is terminated with half-wave open stub to reflect remaining not radiated power back and reuse it. Short-circuited quarter-wave stub also can be used for this purposes. Patches are all equal in size, I did not performed any weighting techniques. Using equal power divider with one arm having addition half-wave phase shift, two series fed sub-arrays can be combined in on corporate feeding array with radiation power maximum at the two center patches. Such series fed arrays can be used as sub-arrays in corporate feeding arrays, or in parallel-series feeding arrays. Prototyping and measurements should be performed.

September 27, 2016: HMC363 prescaler burnt out.

    Few days ago my prescaler stopped to work. HMC363 IC burnt out while doing frequency measurements using small wire probe. It seems that too much power was coupled to prescaler board through the probe. No more readings on frequency meter. Here is the photo of my prescaler board:
HMC363 prescaler

Good news is that new photos will come in better quality, as i repaired my old phone, which contains OV5640 auto-focus camera module.

September 19, 2016:  Some news on VCO project.

    Here is prototype of 5GHz VCO on generic 1.0mm FR4 substrate. Main build blocks are FET transistor, microstrip hairpin resonator and varactor diodes. Few hairpin resonators was manufactured on a PCB without ground plane, so they could be cut out and positioned to match phase oscillation conditions. My idea was to use 5GHz first harmonic to build X-Band VCO which can be used for FMCW sensor. After prototype is done several problems are found. One of the problems is temperature frequency drift. Heating PCB from 25°C to ≈ 80°C leads to ≈ 160 MHz frequency drift at 5GHz, so useful tuning bandwidth is reduced. Varactor diodes must compensate temperature frequency drift and still provide some margin for FMCW modulation. HMC363 divide by 8 prescaler and 2.7GHz frequency counter was used to measure oscillator frequency at 5 and 10GHz. Currently I do not have other measuring equipment. As prescaler was connected through 20dB microstrip coupler, i conclude that 10GHz microstrip band-pass filter performance is acceptable. SIW slot antenna performance is good at 10GHz, more measurements will be performed. Relative dielectric constant at 10GHz of used FR4 material is Er = 4.2. It is confirmed by using 10GHz dielectric resonator positioning. 5200MHz hairpin resonator was replaced by 10500MHz dielectric resonator. The same frequency is obtained each ≈ 7.8mm. It is approximately half wavelength 50Ohm microstrip line at 10500MHz on a 1.0mm FR4 substrate. In parallel feedback configuration DR displacement of half wavelength results in 180 degree shift in both microstrip "arms", which gives total phase shift of one wavelength. I conclude that microstrip resonator oscillator is better to build on Rogers substrate, while dielectric resonator oscillator can be build on generic FR4 substrate. Using Rogers substrate requires to minimize board area, because it affects cost very much. Prototyping on Rogers substrate is also much more expensive compared to FR4 substrate.
Parallel feedback microstrip VCO with hairpin resonator and varactors Microstrip coupler

Laminating machine fix.
    My laminating machine control circuit board was broken, resulting in random LED blinking and constant heating. In my opinion schematic is unnecessary complicated, moreover PCB have two cracked traces and DIP soldering points which peeled off. Thermocouple is potentially exposed to electrostatic discharge from laminating sheet, and it seems that control IC is not protected from such discharge. Mechanical part is pretty good. Motor and heater work directly from 220V AC. I cut out control circuit board. To temporarily fix laminating machine, I connected motor and heater directly to power line. Heating element connected to power line through "on/off" switch. Thermocouple is connected to multimeter. Around three thin A4 sheets could be laminated while multimeter shows resistance of R ≈ 15 kOhm. Then heater must be turned on for around 10 seconds to heat up laminating rollers a bit. At power on around two minutes are required to heat up laminating rollers. This is an A3 laminating machine, so it has pretty large rollers, which cools down slowly.

Laminating machine

August 16, 2016.
    Currently I am working on a proof of concept project of cost-effective microwave sensor system. It must be not only cost-effective, but easy to assemble and require minimal or no tuning. Main parts are VCO, mixer, antenna and DSP circuit. Preliminary layout design of VCO is done already and prototypes performance will be evaluated this month. Experiments on substrate integrated waveguide slotted array antenna (SIW) and patch array antenna will be performed. Another important application of SIW in this project is to realize effective inter-pcb transition of microwave signal, which will allow to use single layer PCB without coaxial interconnections between boards.