A Nanosecond Pulse Generator based on the Reconfigurable Phase-Locked Loop (PLL) Module in Field Programmable Gate Arrays (FPGAs)
Zhang Xing Jian A0226453H, Zhang Jian Ran A0226340R, Chua Rui Ming A0155387U
Field Programmable Gate Arrays (FPGAs) offer both the advantage of Programmable Logic Devices (PLDs) and Application Specific Integrated Circuits (ASICs) because it possesses the functional flexibility without the simplicity of the former and the functional complexity without the inflexibility of the latter1 (Maxfield, 2008). Furthermore, they are cheap and easy to implement which enables small groups to meet their hardware and software needs1 (Maxfield, 2008). As a result, FPGAs represent a useful option to the production of narrow pulses which finds its role in a myriad of scientific applications2 (Zhu & Wang, 2015). In our context, the generation of narrow pulses is of particular interest in the field of integrated photonics, for example, in the carving of pump light into short pulses for Spontaneous Four-Wave Mixing (SFWM) in silicon microring resonators3 (Ma, 2020). The use of FPGAs to produce narrow pulses provides us with a compact and adjustable pulse generator compatible with integrated photonics technology.
Our FPGA is the "DE0-Nano Development and Education Board" from Terasic (vendor)4 (Terasic Technologies, n.d.). The FPGA utilises Intel Altera's Cyclone IV family, specifically, the Cyclone IV E variant5 (Intel, n.d.). It measures 49 x 75.2mm.
Some features of the make-up of our FPGA includes:
- 22,320 Logic Elements
- 594 Embedded memory (Kbits)
- 66 Embedded 18 x 18 multipliers
- 4 General-purpose PLLs
- 53 Maximum FPGA I/O pins
Its memory features 32MB Synchronous Dynamic Random-Access Memory (SDRAM), 2Kb I2C Electrically Erasable Programmable Read-Only Memories (EEPROM), and it has an on-board 50MHz clock oscillator4 (Terasic Technologies, n.d.).
The top and bottom view of the FPGA is featured below respectively4 (Terasic Technologies, n.d.).
Phase-Locked Loop (PLL) Module
The PLL is the key component to our nanosecond pulse generator and would thus be elaborated upon in the following section. A PLL is a feedback control system that locks on the output signal of an oscillator to an input signal by monitoring and maintaining a phase between them. In Altera's FPGAs, the structure of the PLL is as follows7 (Intel, 2018):
- A "Phase Frequency Detector" is used to compare the feedback signal with the input signal for any errors between them
- A "Charge Pump" is used to convert said error, which manifests itself as an error signal, to a correction current
- A "Loop Filter" is used to produce a correction signal from the correction current which is subsequently used to regulate the oscillation frequency of the "Voltage Control Oscillator"
- The signal is then fed through the "Feedback Counter" back to the "Phase Frequency Detector" to complete the loop
Since the PLL is reconfigurable, the phase, frequency and bandwidth of the PLL output signal may be adjusted in real-time. Logic operations may be executed on two signals produced by the PLL to "carve" out the difference between the two signals into a desired narrow pulse. In Zhu & Wang's paper (2015), they utilised the steps and logic operations to obtain such a pulse:
- The production of two clock signals A and B of the same frequency, with a phase shift.
- The implementation of an A OR B operation to produce clock signal C
- The implementation of a D = C XOR A operation to carve out the difference between the two signals to produce the narrow pulse
In our case, our method deviates from that of Zhu & Wang by implementing a two-step procedure to carve out the pulse. This was achieved simply by skipping the A OR B operation. There are two reasons for our choice: Firstly, this simplifies the problem. Secondly, this way, we obtain two pulses for every set of clock signals- At the front-end and tail-end of the first and second clock signals respectively:
- The production of two clock signals A and B of the same frequency, with a phase shift.
- The implementation of an C = B XOR A operation to produce the narrow pulses
For the uninitiated, the OR and XOR logic operation can be described by the truth tables and illustrated by the pulsed operations below as referenced from Electronics Hub8,9 (Electronics Hub, 2017)
FPGA Top Design
The bulk of our efforts went into the code to configure the Nanosecond Pulse Generator. The FPGA Top Design summarises the outcome of our efforts. Our efforts in its entirety could be found here https://drive.google.com/drive/folders/1weGO1Lyi6H_-A1OadQFYIO-wQEv0VXFj?usp=sharing which will lead to a Google Drive where our files are uploaded in its entirety.
FPGA Performance Characterisation
To characterise the performance of our FPGA, we measured the pulses produced by our FPGA with an oscilloscope (see below). We recall that one of the strengths of the FPGA lies in its adjustability; By adjusting the phase difference between the clock signals, and/or the frequency of the clock signals (which can be adjusted in integer multiples of the clock oscillator), we can obtain different performances (in terms of pulse widths and frequencies) of our pulses.
Best Overall Performance
The best overall performance was obtained when the phase difference between the clock signals was set at 11.25 degrees while the frequency of the clock signal is 150MHZ. We obtained a pulse width of 637ps at a frequency of 150MHZ.
Unfortunately, pushing the FPGA further, we were unable to obtain narrower pulses with that higher frequencies.
The best frequency performance was obtained when the phase difference between the clock signals was set at 15 degrees while the frequency of the clock signal is 200MHZ. We obtained a pulse width of 746 ps at a frequency of 200MHZ.
Results as a Function of Phase Difference
We recall from "Methods" that the way to obtain the pulses was to carve the pulses out from two clock signals. Theoretically, the pulse duration should increase with phase difference for the same frequency. When the phase difference between the clock signals were set at 9, 18 and 37.5 degrees and the frequency of the clock signal is 100MHZ, we obtained a pulse width of 805 ps, pulse width of 1.2 ns and pulse width of 2.27ns respectively.
Different Configurations For The Same Pulse Width (Theoretically)
Theoretically, the pulse duration should remain the same if the phase difference and frequency were increased with the same multiple. The plots below feature such a set of measurements. Setting aside some margin of error for the fact that the increase was not in perfect multiples because of the limited step control of the phase difference, as well as the different frequency response of the circuit, this phenomenon was more or less observed.
While we have achieved our goal of creating a working Nanosecond Pulse Generator with our FPGA, it is important to recognise the fact that the pulses created by the FPGA belong to low-power pulse signals between 1.2-3.3V based on the specific I/O standard for the FPGA. Amplification may be required for our FPGA to work with other devices. In this section, a resistance network with a transistor was designed to amplify the output pulses from our FPGA up to 7V for an Electro Optic Modulator (EOM).
As shown in the picture, a TIP29C NPN transistor in our possession was hired as the amplification component. The grounded resistor R4 represents the inner impedance of the EOM. Unfortunately, the 7V half-wave voltage of the EOM required a relatively high current which caused the transistor to overheat. Such issues have to be taken into consideration and represent our future work. Other users would have to consider their own power amplification needs.
Comment: Aaaahhhh.. of you want to have this level shifter working at frequencies of a few 100 MHz, you have to use a transistor that is fast enough to do that. The TIP29 certainly is not, it is designed to work at frequencies around a few kHz only and has a gain-bandwidth product of 3MHz. You would find such information typically in data sheets. For the application you have in mind, you really want to have a wideband transistor with a transit frequency of a few GHz. An example would be a BFR106
 Maxfield, C. (2008). Fpgas: Instant access. In FPGAs: Instant access (1st ed., pp. 1-12). Burlington, MA: Newnes/Elsevier.
 Zhu, Y., & Wang, L. (2015). Design and implementation of nanosecond Pulse generator based ON RECONFIGURATION PLL in FPGA. Proceedings of the 2015 International Conference on Electronic Science and Automation Control. doi:10.2991/esac-15.2015.78
 Ma, C. (2020). High-Quality Photon Pair Generation in Silicon Photonic Microring and Its Applications. UC San Diego. ProQuest ID: Ma_ucsd_0033D_19192. Merritt ID: ark:/13030/m5sf83ns. Retrieved from https://escholarship.org/uc/item/7ws930nj
 Terasic Technologies. (n.d.). All FPGA boards - Cyclone IV - DE0-Nano development and Education Board. Retrieved March 23, 2021, from https://www.terasic.com.tw/cgi-bin/page/archive.pl?Language=English&CategoryNo=139&No=593&PartNo=3#section
 Intel. (n.d.). Cyclone® IV Fpgas devices - Intel® FPGA. Retrieved March 23, 2021, from https://www.intel.sg/content/www/xa/en/products/programmable/fpga/cyclone-iv.html
 Intel. (2020). Intel® Quartus® Prime Software Suite. The Intuitive High-Performance Design Environment. Retrieved March 22, 2021, from https://www.intel.com/content/www/us/en/software/programmable/quartus-prime/overview.html
 Intel. (2018, December 20). Altera phase-locked Loop (ALTERA PLL) ip core user guide. Retrieved March 23, 2021, from https://www.intel.com/content/www/us/en/programmable/documentation/mcn1401782837027.html
 Electronics Hub. (2017, December 24). Digital logic or gate. Retrieved March 29, 2021, from https://www.electronicshub.org/digital-logic-or-gate/#OR_Gate
 Electronics Hub. (2017, December 24). Exclusive or Gate(XOR-Gate). Retrieved March 29, 2021, from https://www.electronicshub.org/exclusive-or-gatexor-gate/#XOR_Gate