# Return to Ritherdon: Photosensitive Epilepsy

This document is to be used as part of the health and safety risk
assessment for the Return to Ritherdon exhibition.

## Summary of Assessment 

The rate of flashes the lighting system produces is not of a higher
enough rate to cause a photosensitive epileptic seizure -- according
to the referenced sources below. Based on the **live test** data, the
system produce a flash rate of **1 Hertz** through out the course of
one day with *approximately* 8 to 9 hours on usage, on the
23/04/2021. This falls outside the **3-30 Hertz** claimed by Epilepsy
Society (link below).

## Information About Photosensitive Epilepsy

According to [Epilepsy
Society](https://epilepsysociety.org.uk/photosensitive-epilepsy), a
flashing/flicker light between 3-30 hertz can trigger a seizure.

> What rate of flashing light can trigger seizures?

> > Between 3-30 hertz (flashes per second) are the common rates to
trigger seizures but this varies from person to person. While some
people are sensitive at frequencies up to 60 hertz, sensitivity under
3 hertz is not common.

More information can be found at:

- [Epilepsy Society](https://epilepsysociety.org.uk/) (home page)
- [Epilepsy
  Action](https://www.epilepsy.org.uk/info/photosensitive-epilepsy)
- [National Health Service](https://www.nhs.uk/conditions/epilepsy/)
  (NHS)

## How This Relates to the Return to Ritherdon Project

The project (unnamed at time of writing) this document refers to is
one artwork in a much bigger project (called Return to
Ritherdon). Within **'unnamed project**, is flashing lights. Because
of that, the rate of flashing/flickering the artwork produces needs to
be reviewed as part of the health and safety risk assessment.

## Viewing the Raw Data

This document will only provide a summarised view of the data because
of the size of it. If you would like to review the raw data please
head to:

- [data](data)
- [raw-data/results](data/results) (for the data after it's been
  processed and used for this risk assessment)

## How the Data was Processed/Reviewed

The data was analysed using the code in the
[Flicker](https://git.abbether.net/return-to-ritherdon/flicker)
repository. Please review the code/repository there for more
information on how the code works -- it is outside the scope of this
document.

## Breakdown of Data Analysis

The sample of data reviewed for this assessment was taken from the
Light Meter readings produced by `factory1` running the software
provided by
[Light-Meter](https://git.abbether.net/return-to-ritherdon/light-meter),
on the **23/04/2021**. Overall, there was **84,294** light readings
taken over the course of **between 8 to 9 hours**. Also, this was a
**live test** with readings taken from the **intended environment**,
under **real-world conditions**. Of those readings, the number of
readings taken is as follows,

*Note: Minutes and Hours are approximate values.*

| Readings-per-Second | Seconds | Minutes | Hours |
|---------------------|---------|---------|-------|
| 1                   | 5,344   | 89      | 1.5   |
| 2                   | 2,955   | 49      | 0.8   |
| 3                   | 13,284  | 221     | 3.5   |
| 4                   | 8,297   | 138.28  | 2     |
|---------------------|---------|---------|-------|
| Total               | 29,880  | 498     | 8.3   |

Let's say you have the following readings (sample taken from
`test-data.csv` stored in [raw-data](raw-data),

**Note: The '.000000' is a artefact from the code's formatting of the
data.** You can ignore it. I've only kept it in to keep the data here
aligned as close as possible with the raw and computed data
[data](data).

| Time-Stamp (YYYY-MM-DD Hr:Min:Sec:MicroSec) | Reading |
|---------------------------------------------|---------|
| 2021-04-23 07:02:50.000000                  | 17      |
| 2021-04-23 07:02:51.000000                  | 17      |
| 2021-04-23 07:02:51.000000                  | 17      |
| 2021-04-23 07:02:52.000000                  | 17      |
| 2021-04-23 07:02:53.000000                  | 17      |
| 2021-04-23 07:02:54.000000                  | 17      |
| 2021-04-23 07:02:55.000000                  | 17      |
| 2021-04-23 07:02:55.000000                  | 17      |

**Note: I've omitted the `Id` column from the example above because is
not relevant for this assessment.** If you review the raw data, you
will come across that column. More often than not, you can ignore it
if you're not looking to review/work with the code written for main
the system.

If you look at the time-stamp `2021-04-23 07:02:51.000000`, you will
see there are two readings recorded. This sample is very small but you
can see the remaining time-stamps have only one reading per each
second intervals. With this in mind, please note there are eight
reading spread across a five-second time-span. That is why the /first/
table above is broken down into time and not just hard frequency
totals. For 5,344 seconds on 23/04/2021, the system took a light meter
reading at one reading-per-second.

According to the sources listed above, 'between 3-30 hertz (flashes
per second) are the common rates to trigger seizures'. This means the
number of readings to review can be reduced to those operating at four
readings-per-second. This is because the lights in this system need to
go from an off-state (1) to an on-state (2) and back to an off-state
(3). The extra (fourth reading) is needed because there needs to be a
starting state (a zero if you will). Below is a sample of a moment
when four readings were taken in a one second period,

| Time-Stamp (YYYY-MM-DD Hr:Min:Sec:MicroSec) | Reading | State |
|---------------------------------------------|---------|-------|
| 2021-04-23 09:04:07.000000                  | 41      | On    |
| 2021-04-23 09:04:07.000000                  | 37      | Off   |
| 2021-04-23 09:04:07.000000                  | 36      | Off   |
| 2021-04-23 09:04:07.000000                  | 36      | Off   |

**Note: The light paired with `factory1` is set to turn on (in the
gallery `gallery1`) when the reading is above 39.**

In the table above, you will see the light changes state (on to off)
once. For it to reach the noted rate of 'between 3-30 hertz', the
third reading in the table will need to have been above `39` to
trigger the light back on. This would have made the light change from
on to off three times, taking into account the need for a 'starting'
state.

Referring back to the table listing the various rates of
readings-per-second (the first table), there are only **8,297**
readings (technically it's seconds but 'readings' is easier to
process) instead of **84,294**. This can be filtered down even more,
though. This is because not all four readings-per-second occurrences
cause the light to change its current state (I.E. it doesn't go from
off to on). With this in mind, the number of entries to review drops
to **8** when you list just the times when the light changes its
state. They are,

| Time-Stamp                 | Readings       | States           | Hertz |
|----------------------------|----------------|------------------|-------|
| 2021-04-23 09:04:07.000000 | 41, 37, 36, 36 | on, on, off, off | 1     |
| 2021-04-23 09:04:11.000000 | 36, 38, 40, 37 | off, on, on, off | 1     |
| 2021-04-23 10:54:05.000000 | 39, 39, 40, 40 | on, on, on, on   | 0     |
| 2021-04-23 10:54:07.000000 | 39, 39, 39, 40 | on, on, on, on   | 0     |
| 2021-04-23 10:56:46.000000 | 40, 40, 39, 39 | on, on, on, on   | 0     |
| 2021-04-23 10:57:13.000000 | 40, 41, 40, 39 | on, on, on, off  | 1     |
| 2021-04-23 10:58:13.000000 | 39, 46, 44, 39 | off, on, on, off | 1     |
| 2021-04-23 11:00:11.000000 | 39, 42, 46, 39 | off, on, on, off | 1     |

This tables shows the limitations of the system to trigger a
photosensitive epileptic seizure. The system produced **at most** a
flash rate of **1 Hertz** over the course of **approximately 8 to 9
hours** of activity.This is below the '3 to 60 Hertz' claimed by
[Epilepsy
Society(https://epilepsysociety.org.uk/photosensitive-epilepsy) (more
references/links at the top of this file).