FSR -> Photodiode latency data analysis

latency_plot.R

library("conflicted")
library("tidyverse")


# Define data sources with file paths and names
data_sources <- list(
  list(
    file_path = "timing_log_simplified_5091.csv",
    name = "iPadM3Pro|Flutter|Comprehensive"
  ),
  list(
    file_path = "timing_log_simplified_9987.csv",
    name = "iPadM3Pro|Swift|Basic"
  ),
  list(
    file_path = "timing_log_simplified_1542.csv",
    name = "iPadM3Pro|Flutter|Basic"
  ),
  list(
    file_path = "timing_log_simplified_7802.csv",
    name = "Pixel7a|Flutter|Comprehensive"
  ),
  list(
    file_path = "timing_log_simplified_3362.csv",
    name = "Pixel7a|Flutter|Basic"
  )
)

# example:

# frame_number,pin_number,pin_state,is_start_pin,trigger_active,pin_states
# 0,2,1,0,0,0;0;0;1
# 322577,0,1,1,1,1;0;0;1
# 324950,0,0,1,0,0;0;0;1
# 325307,1,1,0,0,0;0;1;1
# 327530,1,0,0,1,0;0;0;1
# 534245,0,1,1,1,1;0;0;1
# 537016,0,0,1,0,0;0;0;1
# 537362,1,1,0,0,0;0;1;1
# 539586,1,0,0,1,0;0;0;1
# 554218,0,1,1,1,1;0;0;1
# 557208,1,1,0,0,1;0;1;1
# 557309,0,0,1,0,0;0;1;1


# Function to find previous FSR event for each photodiode event
calculate_latencies <- function(fsr_events, photodiode_events) {
  latencies <- numeric(nrow(photodiode_events))
  
  for (i in 1:nrow(photodiode_events)) {
    photo_frame <- photodiode_events$frame_number[i]
    
    # Find the most recent FSR event before this photodiode event
    previous_fsr <- fsr_events %>%
      dplyr::filter(frame_number < photo_frame) %>%
      slice_max(frame_number, n = 1)
    
    if (nrow(previous_fsr) > 0) {
      latencies[i] <- photo_frame - previous_fsr$frame_number
    } else {
      latencies[i] <- NA
    }
  }
  
  return(latencies)
}

# Function to process a single data file
process_data_file <- function(file_path, data_name) {
  cat("Processing:", data_name, "\n")
  
  # Read the data
  bela_data <- read_csv(file_path, show_col_types = FALSE)
  
  # Filter FSR events (input events)
  fsr_events <- bela_data %>%
    dplyr::filter(pin_number == 0, is_start_pin == 1, trigger_active == 1) %>%
    arrange(frame_number)
  
  # Filter photodiode events (display events)
  photodiode_events <- bela_data %>%
    dplyr::filter(pin_number == 1, is_start_pin == 0, trigger_active == 1) %>%
    arrange(frame_number)
  
  # Calculate latencies in frames
  latency_frames <- calculate_latencies(fsr_events, photodiode_events)
  
  # Convert to milliseconds (44.1 kHz = 44100 samples per second)
  sampling_rate <- 44100
  latency_ms <- latency_frames / sampling_rate * 1000
  
  # Remove NA values and create data frame
  valid_latencies <- latency_ms[!is.na(latency_ms)]
  
  if (length(valid_latencies) > 0) {
    return(data.frame(
      latency_ms = valid_latencies,
      source = data_name
    ))
  } else {
    cat("Warning: No valid latencies found for", data_name, "\n")
    return(data.frame(latency_ms = numeric(0), source = character(0)))
  }
}

# Process all data files
all_latency_data <- map_dfr(data_sources, function(source) {
  process_data_file(source$file_path, source$name)
})


summary_stats <- all_latency_data %>%
  group_by(source) %>%
  summarise(
    n_events = n(),
    mean_ms = round(mean(latency_ms), 2),
    median_ms = round(median(latency_ms), 2),
    min_ms = round(min(latency_ms), 2),
    max_ms = round(max(latency_ms), 2),
    sd_ms = round(sd(latency_ms), 2),
    .groups = 'drop'
  )
# order sources by lowest to highest mean latency
summary_stats <- summary_stats %>%
  arrange(mean_ms)

# set the levels of the source factor to make sure that the lowest latency source is first
all_latency_data$source <- factor(all_latency_data$source, levels = summary_stats$source)

# Create side-by-side violin plot
latency_plot <- ggplot(all_latency_data, aes(x = source, y = latency_ms, fill = source)) +
  geom_violin(alpha = 0.7) +
  geom_boxplot(width = 0.2, fill = "white", alpha = 0.8) +
  labs(
    title = "Display Latency Distribution Comparison",
    subtitle = "FSR to Photodiode Response Time",
    x = "Data Source",
    y = "Latency (ms)"
  ) +
  theme_minimal() +
  theme(
    axis.text.x = element_text(angle = 45, hjust = 1),
    plot.title = element_text(hjust = 0.5),
    plot.subtitle = element_text(hjust = 0.5),
    legend.position = "none"
  ) +
  scale_fill_viridis_d() +
  scale_y_continuous(breaks = seq(0, max(all_latency_data$latency_ms, na.rm = TRUE), by = 5))

# Print summary statistics for each source
cat("\n=== LATENCY SUMMARY BY SOURCE ===\n")

print(summary_stats)

# Display the plot
print(latency_plot)