Grape Version 1: First prototype of the low-cost personal space weather station receiver

Graphical abstract

by less than typical wavelengths of MSTIDs) and cadences of measurements shorter than 5 min. The detection of MSTIDs with the system described in this paper is a target of current research. Improved TID detection will, in turn, support the exploration of frontier questions in atmosphere-ionosphere coupling by helping to establish the relationships between TIDs and acoustic gravity waves (AGWs).
Today, with the advantage of cheaper instrumentation and computational power, synchronized HF Doppler measurements can be conducted by many stations simultaneously. This advance is due not only to the availability of inexpensive single-board computers such as the Raspberry Pi, but also to the advent of inexpensive GPS-disciplined oscillators. Taken together, fleets of individual stations can be considered as a single meta-instrument (i.e., an instrument made of contributions from many individual instruments), capable of observing traveling ionospheric disturbances and other ionospheric perturbations on a continental or global scale.
The first light of the citizen science meta-instrument was a pilot experiment in October 2019, the Festival of Frequency Measurement, held to celebrate WWV's centennial. In this one-day campaign, participants across the United States used their amateur radio equipment to gather frequency estimation data, as described in [8]. While that study demonstrated a high correlation between nearby stations and showed promising results despite the variety of experimental apparatus used, it also showed its limits: the volunteers' equipment could not be used for ordinary activity during the experiment, which would be prohibitive for long-term data collection; the varied signal chains resulted in errors which required some datasets to be culled; and the need for shortwave receivers costing hundreds or thousands of dollars created a barrier to entry that restricted the scope of the data collection to members of the amateur radio and shortwave listening communities, despite the fact that most features of these receivers are surplus to our needs for this particular experiment.
To address these limitations, we present the Grape Version 1, a low-cost, purpose-built low IF receiver for Doppler observations on 2.5, 5 and 10 MHz. The Grape Version 1 will be succeeded by the Grape Version 2, which will include additional channels for simultaneous monitoring of other frequencies of WWV (15 MHz) and CHU (3.33 MHz, 7.85 MHz and 14.67 MHz).
Nota bene: Where possible, amateur radio callsigns are used herein, in addition to names, in order to specify individuals and club stations. Because these callsigns are unique and persistent identifiers, they support the Findability criterion of FAIR Data principles. The authors' callsigns are N8OBJ, KD8OXT, AD8Y, and W2NAF, respectively.

Hardware description
The Grape is a very simple heterodyne receiver, as illustrated in Fig. 3. The GPSDO signal acts as the local oscillator, whose signal, set to 1 kHz below the carrier frequency of interest, is mixed with the filtered input signal. The output of this operation is an intermediate frequency at the difference frequency of 1.000 kHz. This is the signal that is used to measure the Doppler shift of the carrier frequency.
In the use case it is designed for, its primary advantage is low cost. Most of the system cost derives from the GPSDO and the Raspberry Pi, rather than the mixer board. The cost of the mixer board itself is about $30, and it stands in for a transceiver which may cost hundreds or thousands of dollars, e.g., the Icom 7610. There are few communications receivers which provide for high-accuracy frequency references such as a GPSDO, and the ones that do are expensive. Additionally, the Grape's signal chain is simple and easily characterized, as opposed to the ''black box" posed by the more convoluted signal chain of a conventional transceiver. The Grape is an analog receiver and will have no more time delay incurred than that from its filters. Most contemporary communications receivers use digital filtering with phase and group delays that distort measurements for the Grape's purposes.
There are additional aspects of the Grape which make it more suitable for these measurements than most commercial receivers. Very few communications receivers have provisions for shutting off automatic gain control (AGC), which makes absolute received signal level measurements impossible. The Grape has no AGC; at risk of overload, it can act as a field strength measuring device. Frequency accuracy is a function of the GPSDO rather than of the Grape. Grape selectivity will not be expected to be as high as that of an excellent communications receiver, but there are few adjacent-channel interfering signals on the frequencies of interest in this use case.
The Grape is designed to monitor 2.5, 5, or 10 MHz, but could be customized to another frequency in the HF band by modifying the inductance and/or capacitance values of the input filter. In use cases where the precision of a GPSDO is not required, an inexpensive oscillator circuit can be substituted for the local oscillator. Such uses include relative time of flight measurements on the transmitter's modulations such as the second ticks. Purpose-designed modulations such as those from the WWV/H Scientific Modulation Working Group [9] also fall into this category. It can also be used as the basis for a simple classroom demonstration of a heterodyning receiver.
The open-source software accompanying the Grape generates daily records of received signal amplitude and frequency shift, in the form of both plots and.CSV files.
Inside and outside of the original user community, some uses for the Grape V1 include: Constant monitoring on single frequency, with modifications to the board as described above. Alerts can be set if the amplitude or frequency of the signal goes outside an expected range. In the amateur radio space, it could be modified to monitor the band edge, and indicate transmissions outside allocated spectrum. Propagation studies. The Grape is small and highly portable, and could be used to check for presence of a beacon signal on a particular frequency at a given location, provided that antennas are also consistent. An example of a Grape packaged for portability is shown in Fig. 4.  Carrier dropout monitoring. In addition to checking for the absence of a carrier, the Grape could be used to check for the outage of a carrier in a particular sequence as a form of communication, as was used in the CONELRAD system in the 1950s. Educational applications. The Grape is a good introductory surface-mount soldering kit, appropriate to high school or college level, and can be deployed as a long-term classroom experiment to stimulate interest in space weather.

Project components
In addition to the mixer board, a functional PSWS requires a tunable GPS-disciplined oscillator, a central computer, and an HF antenna. A monitor, mouse and keyboard will also be required to interface with the Raspberry Pi. The GPSDO can be programmed from an external computer.

Electronic components
The total cost of electronic components for the board is $22.49, as listed in Table 3.

Mixer board assembly
The unpopulated circuit board may be purchased from OSHPark, as noted in Table 2, and populated with the parts listed in Table 3. The printed circuit board is shown in Fig. 6, and the populated board is shown in Fig. 7.
A few recommendations for hand assembly: First, the reader is advised to purchase extra components in case of loss or damage. Assembling the board in a tray or cookie sheet with low sides may be helpful for keeping track of parts. One should use a soldering iron with a small tip (e.g., the Hakko FX888D-23BY Digital Soldering Station with the T18-BR02 tip) and keep the tip clean. When soldering surface mount parts, it is easiest to put solder on one pad then mount the part, soldering that pad first in order to adhere the part to the board, then soldering other pins. Hold each part gently with tweezers when placing it on the board. Using a small diameter spool of solder makes assembly easier as well.  The orientation of the mixer IC (U1) is indicated by a small white circle on the PCB near pin 1. The package will only identify the orientation by a slight slant on one edge of the package, which indicates which side pin 1 is on. It should be oriented with the slant toward the outside of the PCB. Looking from the top of the IC with the slant on the left side, pin 1 is on the top left and numbers progress counter-clockwise around the package (like any standard DIP IC package).

Mixer board testing
Once the board is assembled, it can be bench tested with a Leo Bodnar GPSDO, RF signal generator, power supply and oscilloscope. First, apply 6-18 V DC to the VIN power pins and verify that the power indicator LED turns on. At this point If all that checks out, it is time to do a functional test. To do this, connect the Leo Bodnar GPSDO to the LO input on the board and set the frequency to 9.999 MHz with an output drive current of 8 mA.
Set the J1 JMPR on the Grape PCB to the 10 position and connect a RF signal generator to the ANT input jack. Set the output frequency to 10.000 MHz and the output level to 100 lV peak (not RMS). Next, connect the oscilloscope to the audio output VOUT (pin 1 is the signal output, pin 2 is ground). On the oscilloscope, you should observe a 1 kHz sine wave with an amplitude of approximately 75-100 mV (depending on the skew of the LC part values in the front end filter). If you don't have access to an oscilloscope, headphones can be used to verify the presence of a 1 kHz tone.

System assembly
Once the mixer board has been assembled and tested, the system should be assembled as shown in Fig. 8: The USB audio adapter is plugged into the Raspberry Pi (B). Facing the Raspberry Pi USB connectors, put it into the top right position. The audio cable is connected to the pink connector input of the USB audio adapter coming from the VOUT audio output of the mixer board. The power pins on the mixer board are wired to the 5 V pins on the Raspberry Pi. (C) Alternatively, the board can be powered externally (6-18 V DC) as shown in the construction in Fig. 9. The GPSDO is powered from the Raspberry Pi using the included USB cable. (D) Plug the GPSDO USB cable into the lower left USB connector on the Raspberry Pi (diagonally across from the audio USB adapter). The GPS antenna (included with the Bodnar) is connected to the GPSDO. The antenna itself is not shown in Fig. 8.

(E)
The antenna input on the Raspberry Pi is connected to an appropriate HF antenna using an SMA -SO239 cable (F). The receive antenna (assumed to use a PL-259 connector) is further discussed in Section 5.4.  The Leo Bodnar GPSDO is powered from the Raspberry Pi over USB, and draws 250 mA at 5 V. The mixer board draws 7.2 mA and can be wired to the 5 V pins on the Raspberry Pi (6-18 V DC is preferred, but 5 V will work). Thus, the primary 5 V power connection for the system is the power supply for the Raspberry Pi, and the total power draw is anywhere from 250 mA to 1.3 A, depending on what is running on the Raspberry Pi.

Receive antenna
Due to the variety of suitable HF antennas, detailed discussion of the receive antenna is outside the scope of this paper. We refer the reader to external resources, such as the ARRL Antenna Book [10]. Versatile HF antenna kits are also available. Alternatively, a long wire antenna may be used with a balun, such as a NooElec Balun One Nine, available on Amazon for $11.95. In this case, the SMA to SO-239 in Table 2 is replaced with a second SMA-M to SMA-M cable.
It is critical that the antenna be properly grounded to a real earth ground. For a wire antenna connected to one side of a balun, ensure that the other balun input is connected to earth ground. Poor grounding will deprive the system of a reference point, so the signal will be noisy and weak.

Software installation
The OS image for the Raspberry Pi is available on Mendeley with the design files. Burn this image to the micro SD card using balenaEtcher or a similar program, then insert the card into the Raspberry Pi and boot it. Once the Pi is running, follow the instructions provided with the OS image file to complete the installation. Detailed instructions are provided in MakeGra-pe1OSImage.pdf. This image includes a version of the open-source software fldigi 4.1.13 which we have modified to provide file labeling and metadata suitable for this project.

Operation instructions
Once the software is installed and running, it will generate daily plots of the filtered signal in.PNG format and raw data in. CSV format. An example of one such plot is shown in Fig. 10.
Each day, when the operating system time hits 00:00:00 UTC, the customized version of FLDigi 4.1.13 begins running on this node to start a new day's data collection file. This data file is stored in /PSWS/Srawdata/ as file analysisYYMMDD.csv, where YYMMDD is today's new UTC date. The program then determines the beacon being monitored and checks and updates (if needed) the associated /PSWS/Sinfo/Beacon1 file (associated with the station Radio1) with the beacon identifier string (''WWV5", for example). The program also reads system-related information (Node Number, Gridsquare, Latitude/ Longitude/Elevation, City State, RadioID, Beacon1) from this same /Sinfo directory. This is done to create the first line header info to aid in the plotting of this data file in the future (initially next day). If the frequency being monitored doesn't match Fig. 10. Example of daily plot, spanning 24 h from midnight to midnight UTC on 24 June 2020, with sunrise peak highlighted. The black line indicates the frequency shift of the carrier, and the red line indicates the amplitude of the signal. In Cleveland, where this data was taken, peaks indicative of traveling ionospheric disturbances (TIDs) are visible in the nighttime half of the plot, and the carrier signal becomes weak and noisy after sunrise. A plot like this is automatically generated by the Grape software each day. The data represented in this and other plots are archived at [11]. any of the 9 known beacons, it declares it as an 'unknown' beacon and is saved/plotted accordingly (with filename support indicating this).
Data collection continues in this file for the entire UTC day. If the data collection process is interrupted or stopped, the data is saved. When the process restarts, the new data gets appended to this same analysisYYMMDD.csv file where the data collection was interrupted as long as the UTC date is still the same. This allows for system down time, maintenance, system crashes, etc. The beacon frequency data is created and stored at a one second cadence to the.csv file by the data collection program. This gives 86,400 data entries per day in this file.
To pause data collection during the course of the day-for example, if one wishes to use the antenna for radio operationthe fldigi program may be set to NULL mode and the antenna disconnected. Afterwards, reconnecting the antenna and setting fldigi to Freq Analysis mode will allow data collection to be seamlessly resumed.
Data collection continues throughout the day until the system clock hits 00:00:00 UTC again, and the process restarts. Also at 00:00:00 UTC, a crontab job for the user pi kicks off a python program to check the data file for corruption and to rename, process and move a new version to the correct directory for submission to the central node server. If the frequency directory does not exist yet for the newly processed frequency file, it is created (with correct file permissions 664) and the data then stored there. If the beacon is not in the list of known beacons, it's named 'unknown' and stored as such. The script also removes the Freq Err and dB(Vpk) data columns from the file to be submitted to the server as well (these numbers can be derived from the data already there and are therefore considered redundant). The original analysisYYMMDD.csv is left in the /Srawdata/ directory for future access and for plotting. If the filecheck program finds any problems, it fixes and stores them (in /Srawdata) and appends the .bad file extension to the original file. It is kept for future reference.

Validation and characterization
Once the system is running, some observation is required in order to verify that the resulting signals are geophysical in nature, rather than the result of local RF interference. As noted in Section 5.4, a proper connection to earth ground is imperative.
An example of local interference is shown in Fig. 11: the relatively flat red signal strength line in the top two plots show where leakage from a local standard and connected equipment is stronger than the signal from WWV. Even when the local signal is not stronger than WWV, a comparison of the top two plots with interference present and the bottom two interference free plots suggests a significant reduction in the measured amplitude of the Doppler shift received from WWV.
To reduce this interference, one can use standard techniques: better shielding, breaking ground loops, ferrite chokes, careful attention to balanced/unbalanced transformation, etc., or moving the antenna farther from the 10 MHz source. A transmission line common mode choke may also be used to minimize pickup of the local signal by the outer shield of the coax going to the WWV receiver. Double shielded coax between a 10 MHz standard and other equipment may be required. Fig. 11. Four daily plots, recorded on consecutive days, from a station in Arizona. The first two show local interference from a 10 MHz receiver. Image credit: Joe Hobart W7LUX.
One key indicator of a working system is the distinctive peak in the frequency plot shortly after local sunrise, when solar input to the ionosphere causes the virtual height to drop. This is a signature of the ionospheric diurnal variation. The ionosphere is formed primarily by photoionization of the neutral atmosphere due to solar ultraviolet (UV) energy. During the day, ionospheric electron densities increase due to photoionization; at night, ionospheric electron densities decrease due to recombination [12]. During the dawn transition, the increase in electron density causes a shortening of the propagation path with time, producing the positive Doppler sunrise peak signature observed between 10 and 12 UT on Fig. 10. As the ionosphere reaches daytime equilibrium, the Doppler shift returns to near-zero values for the rest of the day.
As indicated by the title block, the data in Fig. 10 was taken by station AD8Y in Cleveland, Ohio on June 24, 2020. The plot shows the frequency and amplitude of WWV's carrier on 5 MHz over the course of a UTC day. An annotation has been added to indicate the sunrise peak. Plots from the same week are shown in Fig. 12: despite significant variation throughout the day, each plot shows a sunrise peak. Over time, the data shows solar cycle variations, seasonal variations, and movement of the sunrise with the changing length of day.
The best verification method is to deploy two stations in the same local area and compare data. Fig. 13 shows plots made from WWV's 5 MHz carrier from two stations about eight miles apart. Shared features in the frequency plots, like the distinct peaks between 03:00 and 05:00 UTC, are a strong indicator of geophysical variation.
Capabilities of the Grape Version 1: Fig. 12. Plots from preceding (top row) and succeeding (bottom row) days for comparison to Fig. 10, showing the range of daily variation. Note that a sunrise peak is visible in each plot. Low-cost monitoring of WWV and WWVH on 2.5, 5, and 10 MHz Daily plotting to facilitate scientific engagement.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.