destevez – Daniel Estévez

Multispectral analysis of the Tres Cantos wildfire with Sentinel-2 data

A few days ago, I posted about a recent wildfire in Tres Cantos, Madrid, Spain, sharing images of Pléiades Neo and Sentinel-2 that showed the extent of the fire. Since the Sentinel-2 data products can be downloaded for free and they include multispectral data covering 13 bands including visible, near infrared and short wave infrared spectrum, I have decided to do an analysis of this data. There are two main goals to this. The first is to learn. I have never done this before. It is an interesting topic, and I mostly learn by doing. So bare with me that although all my results seem to make sense, I could have made some mistakes or done something in a suboptimal way.

The second is to understand how badly the holm oak woodland in “Monte de Viñuelas” has been burnt. In the previous post I explained that this is a 30 km² estate and that a large part of it has been hit by the fire. From the wall that borders the estate I can see that the oaks that are near the wall have not burnt, but the grass in the ground has burnt completely. The leaves on these oaks are mostly fine. Some of them might wither with time if the fire has killed the tree (in some previous smaller fires I have seen that if the fire has killed only part of the tree, just that part will wither). Here are some images published by a hunting journal showing how some of these areas look like a few days after the fire.

However, deeper into the woodlands there seems to be more damage and trees burnt completely. This is not so easy to see from the wall, and I cannot just walk in, as it is private land. Aerial photography would be best, but without it, satellite imagery is the next best thing. It is hard to tell how large the damage is from the visible spectrum imagery, but as we will see in this post, the infrared bands have much more information.

Tres Cantos wildfire

This post is going to be slightly unusual for the topics of this blog, because there is no RF, but nevertheless there is space-based remote sensing, which fits somewhat well with the things I usually write about. I wanted to write down this information somewhere, and it was too long for a series of tweets.

As some of you might have heard in the news, there has been a large wildfire in Tres Cantos, Madrid, Spain, which is the city where I live. This has even been featured in international news. First of all, I am okay, as are all the family and friends I have in the city. The fire broke out on 2025-08-11 17:45 UTC (19:45 local time) and by the next morning its perimeter was already contained. As of writing this post on the morning of 2025-08-13, the fire is almost put out and is considered to be controlled. We have been lucky that a fire so close to the city has caused relatively low damage. I am not keeping a tally, but what I heard is: one person’s life, a few houses in the borders of the city, as well as a few countryside houses and sheds, the King’s College British school, and the 17th century Viñuelas castle, as well as part of the castle grounds, which consist of 3000 hectares of holm oak woodland, commonly known as “Monte de Viñuelas”.

Since I woke up on the morning of 2025-08-12, I have been very interested in understanding which area has been affected by the fire. The information I could see in Google maps, and even in some news articles (which could have been based off Google maps) didn’t quite match what I had seen in pictures and videos shared in social media, as well as what I saw by driving on the streets bordering the town. An official map has not been published, as far as I know. So I have been keeping an eye on space-based imagery platforms to see when the first images taken on 2025-08-12 would pop up. I don’t use these services frequently, so this has also helped me get up to speed on the current constellations, platforms and services. This is the topic of this post.

n78 band 5G NR recordings

In my last post about 5G NR, which was part of a series in which I analyzed the signals in a short recording of an idle srsRAN gNB, I mentioned that I had already decoded all the signals that appear in the recording, and that to move on with my 5G series I would need to make and use some more complex real world recordings next.

A 5G band I’m particularly interested in is n78 (3.3 – 3.8 GHz TDD). This is being used to deploy 5G in many European countries, including Spain, as showed by this list in Wikipedia. Due to the large bandwidth available, it is common to see cells with 100 MHz bandwidth in this band.

Galileo OSNMA chain renewal

Galileo OSNMA (open service navigation message authentication) is a cryptographic system that is used to authenticate the navigation message (satellite ephemeris and clocks, etc.) in the Galileo GNSS. I have spoken before about OSNMA in this blog, since I implemented an OSNMA library in Rust a few years back. A good introduction to OSNMA for readers unfamiliar with how it works can be found in Bert Hubert‘s short series of OSNMA posts. The OSNMA system is currently in the public observation test phase.

On July 4, an OSNMA live test notification went out with the following message:

EVENT DESCRIPTION: USERS ARE ADVISED THAT, AS PART OF THE PUBLIC OBSERVATION TEST PHASE ACTIVITIES, A TESLA CHAIN RENEWAL IS PLANNED ON 2025-07-07 10:00 UTC AND THE TRANSITION WILL OCCUR ON 2025-07-08 10:00 UTC. THE TESLA CHAIN RENEWAL PROCESS IS DESCRIBED IN THE OSNMA SIS ICD (LINK).

NOTE THAT USER RECEIVERS SHALL PREVENT THE USE OF ANY CHAIN THAT HAS BEEN SUBJECT TO A RENEWAL PROCESS.

I have used the utilities from the Galmon project to record the Galileo INAV data received by a uBlox GNSS receiver that I have at home. This dataset can be used to test OSNMA implementations and to study how the chain renewal was done. The dataset is publised in Zenodo as “Galileo INAV data for OSNMA chain renewal test in July 2025“. In this post I study the chain renewal using my galileo-osnma Rust implementation.

About FLLs with band-edge filters

Using band-edge filters for carrier frequency recovery with an FLL is an interesting technique that has been studied by fred harris and others. Usually this technique is presented for root-raised cosine waveforms, and in this post I will limit myself to this case. The intuitive idea of a band-edge FLL is to use two filters to measure the power in the band edges of the signal (the portion of the spectrum where the RRC frequency response rolls off). If there is zero frequency error, the powers will be equal. If there is some frequency error, the signal will have more “mass” in one of the two filters, so the power difference can be used as an error discriminant to drive an FLL.

The band-edge FLL is presented briefly in Section 13.4.2 of fred harris’ Multirate Signal Processing for Communication Systems book. Additionally, fred also gave a talk at GRCon 2017 that was mainly focused on how band-edge filters can also be used for symbol timing recovery, but the talk also goes through the basics of using them for carrier frequency recovery. Some papers that are referenced in this talk are fred harris, Elettra Venosa, Xiaofei Chen, Chris Dick, Band Edge Filters Perform Non Data-Aided Carrier and Timing Synchronization of Software Defined Radio QAM Receivers and fred harris, Band Edge Filters: Characteristics and Performance in Carrier and Symbol Synchronization.

Recently I was looking into band-edge FLLs and noticed some problems with the implementation of the FLL Band-Edge block in GNU Radio. In this post I go through a self-contained analysis of some of the relevant math. The post is in part intended as background information for a pull request to get these problems fixed, but it can also be useful as a guideline for implementing a band-edge FLL outside of GNU Radio.

Z-Sat VHF transmissions

Z-Sat is a microsatellite by Mitsubishi Heavy Industries that was launched in 2021. It is a demonstrator for multi-wavelength infrared Earth observation technologies. It carries an amateur radio payload that was coordinated by IARU and which consists of a BBS (bulletin board system) with a 145.875 MHz downlink and 435.480 MHz uplink. I have not been able to find more information about the amateur radio payload on this satellite.

Recently, Daniel Ekman SA2KNG asked me to analyze some transmissions by this satellite. Apparently it has recently begun to transmit a digital modulation, as shown in this SatNOGS observation, while it typically had transmitted CW telemetry in the past. The point where this started appears to be on 2025-06-20, as there is a SatNOGS observation of CW telemetry on that day followed by an observation of the digital modulation. In this post I analyze this digital modulation and explain what it is.

5G NR PDSCH

In my previous post in the 5G NR RAN series, I showed how to decode the PDCCH (physical downlink control channel), which is used to send control information from the gNB (base station) to the UEs (cellphones). In this series I am using as an example a short recording of the downlink of an srsRAN gNB. The PDCCH transmission that I decoded in the previous post was a DCI (downlink control information) containing the scheduling of the SIB1 PDSCH transmission. The PDSCH is the physical downlink shared channel, which is the channel where the gNB transmits data. The SIB1 is the system information block 1. It contains basic information about the cell, and it is decoded by the UE after decoding the MIB in the PBCH, as part of the cell attach procedure. In this post I will show how to decode this PDSCH SIB1 transmission.

5G NR PDCCH

This is a new post in my series about the 5G NR RAN. As in previous posts, I am analyzing a short recording of the downlink of an srsRAN gNB. There are no UEs connected to the cell during this recording, so there isn’t much interesting traffic, but the recording contains all the essential 5G signalling. In particular, there is a SIB1 transmission in the PDSCH, with its corresponding transmission in the PDCCH.

The PDCCH (physical downlink control channel) is used to transmit control information to the UEs in the form of DCI messages (downlink control information). The most common types of DCIs are those that specify the scheduling parameters of transmissions in the PDSCH (physical downlink shared channel), and the uplink grants for UEs in the PUSCH (physical uplink shared channel). The role that the 5G PDCCH plays is very similar to the role that it plays in LTE, so my post about the LTE PDCCH can be good for more context. However, in 5G the channel coding and physical layer of the PDCCH is substantially different from LTE.

Time-dependent delay in GNU Radio

Since some years ago, a Doppler correction block has been available in gr-satellites. This block uses a text file that defines a time series of frequency versus time, and applies the appropriate frequency shift to each sample by linearly interpolating the frequency corresponding to the time of that particular sample. It can be used both to correct Doppler in a satellite propagation channel and other similar channels, as well as to simulate Doppler.

For some time I have been wanting to implement a similar block that applies a time-dependent delay to a complex baseband signal. The delay versus time would be defined by a text file in the same way as for the Doppler correction block. With this block, the delay of a satellite propagation channel can be simulated. It can also be used to correct the delay-rate of a satellite channel, which causes waveforms to be expanded or compressed in time due to a changing time-of-flight. This effect is known as code Doppler in GNSS, and as symbol frequency offset in general digital communications.

For accurate simulation of a satellite propagation channel, both the Doppler block and the delay block are needed, since the Doppler block accounts for the effects of variable time-of-flight on the RF carrier, and the delay block accounts for the effects on the band-limited modulation of the signal, by delaying its complex baseband representation.

I have now added a Time-dependent Delay block to gr-satellites. In this post I give a few details about its implementation and usage.

Analysis of the CAMRAS Venus radar experiment

On March 22, CAMRAS performed a Venus radar experiment (or Earth-Venus-Earth bounce, as it is more commonly known in Amateur radio) in collaboration with Astropeiler Stockert, the Deep Space Exploration Society, and the Open Research Institute. The experiment was done in L-band, at 1299.5 MHz, using the 25 m Dwingeloo radiotelescope as transmitter and receiver and the Stockert 25 m telescope as a receiver. The experiment was done during the Venus conjunction, as this minimizes the distance between the Earth and Venus. The round-trip time to Venus was approximately 280 seconds. The radar waveform was a CW carrier with a duration of 278 seconds. It was transmitted a total of 4 times every 600 seconds. Thus each period was composed by:

278 second transmission
~2 second TX/RX guard time
278 second echo reception
~42 second wait

More transmissions were planned, including some spread-spectrum signals designed by the Open Research Institute. However, the transmitter started failing and they had to stop.

Update 2025-04-21: I have been informed that the spread-spectrum signal for this experiment was designed by CAMRAS. The waveform that the Open Research Institute is designing will potentially be used in future experiments.

CAMRAS has published an analysis of the recorded IQ data, showing successful echo detections with Dwingeloo and Stockert. There is an example Jupyter notebook that shows how to Doppler correct the echo and detect it with an FFT. All the recorded IQ data has been published.

In this post I will do my own analysis of the experiment. The goal is not to confirm the successful detection with an independent analysis, since I believe that the analysis published by CAMRAS is correct and leaves no doubt about this. It is to expand this analysis and to touch on other topics that this analysis has not covered:

Calculation of the Doppler. CAMRAS has published CSV files containing the expected Doppler at each receiver, but they have not published the code to calculate this. Here I will do all the relevant calculations with SPICE.
Doppler correction with the gr-satellites Doppler correction block, which performs linear interpolation of the Doppler frequency to calculate the frequency shift applied to each sample.
High-quality spectral analysis with a polyphase filterbank.
Try to estimate the Doppler spread and compare with some results in the literature.