NEVER FLY IN FOG

[monty.roman]

bottom IR sensors are easy to get confused. The only way to descend through the fog safely (not considering humidity which is dangerous for any unprotected electronic device) is to disconnect bottom sensors. Other drones like Mavics have an option to disable Landing protection which is pretty much it.

Some guys from our CZ Spark community tried to disable sensors by modifying firmware and they partially succeeded (they turned them off but they were not able to turn them back on, even by reflashing fresh FW).

Some people even tried to cover sensors by duct tape but forgetting that infrared sensors are able to see through the tape partially making them more unreliable than before.

 

[ChristoffB]

First hand experience and expensive lesson learned - was photographing the fog coming in from the ocean. It was coming in quicker than I thought. I saw the danger and started to descend, albeit too late. Got message "Infrared Sensor Failure" and the aircraft initiated landing. Came down at full speed. Could not cancel the process. My reactions were to slow and it clipped a fence and crash landed, resulting in a cracked camera lens.

So call it BS if you like and take the risk. I will not do it again in the future.

 

[LKLM-Media]

First hand experience and expensive lesson learned - was photographing the fog coming in from the ocean. It was coming in quicker than I thought. I saw the danger and started to descend, albeit too late. Got message "Infrared Sensor Failure" and the aircraft initiated landing. Came down at full speed. Could not cancel the process. My reactions were to slow and it clipped a fence and crash landed, resulting in a cracked camera lens.

So call it BS if you like and take the risk. I will not do it again in the future.

I think regardless of the messages we are seeing - (all of which seem to be slightly different) - and slightly different symptoms (yours ascended quickly, whereas mine said it was doing so, but instead just hovered because it also kept thinking it had reached the ground...)
The reality is - the IR sensors are there, and they Do Stuff! Not always what we expect...But mostly it is to design! I wanted to test something using Litchi today - and almost decapitated myself!
I had a Lenovo tablet I have purchased for nothing other than DJI work...I had it leaning on the front of the car (no mount for it yet.) but had the awesome Speech giving me feedback on battery, runtime, height, etc. - I brought the drone down and was bringing it towards me, but wanted to quickly see what was on the screen (it was getting dark and I couldn't tell if the Gimbal was forward or down from the previous position while ascending.
Next thing I know I looked up and Spark was inches from my face - I don't know if Litchi handles the Proximity differently or not...It did start beeping a moment later when I moved my body back in front of it - But I wondered if it was because previously when using the DJI GO 4 app on the Phone I had turned avoidance off. In either case - Inches vs the normal 2.5m surprised me.

If anyone wants a laugh - here is a link to my almost unfortunate demise!

 

[LKLM-Media]

bottom IR sensors are easy to get confused. The only way to descend through the fog safely (not considering humidity which is dangerous for any unprotected electronic device) is to disconnect bottom sensors. Other drones like Mavics have an option to disable Landing protection which is pretty much it.

Some guys from our CZ Spark community tried to disable sensors by modifying firmware and they partially succeeded (they turned them off but they were not able to turn them back on, even by reflashing fresh FW).

Some people even tried to cover sensors by duct tape but forgetting that infrared sensors are able to see through the tape partially making them more unreliable than before.

I'm curious as to whether the Bottom camera is nothing more than a CCD Sensor which picks up the IR bounced from the two IR "holes" - or whether they are independant, and the 2 sensors are send and receive, and the Camera has a bit of Object detection written into it - fairly easy to do even with small CPU usage...

One more thing I found out in Prodding around, apparently the Code used for the Spark is significantly the same in many parts to that used in the Mavic - which has Ultrasonic Sensors, and for simplification the term "Utrasonic Sensors" which some people are seeing - still relate to the IR sensors on the Spark. Just named to confuse you!

The problem with putting any tape on - as you said is the signal might still go through - the in reality if each of the sensors is both a send and receive - Putting anything over it - much like the Fog itself, will cause an immediate bounce back - you would need instead to use an IR absorbing material or something which Scatters the IR beam so that it had no way to return to the Sensors.

 

[Spark 317]

I'm curious as to whether the Bottom camera is nothing more than a CCD Sensor which picks up the IR bounced from the two IR "holes" - or whether they are independant, and the 2 sensors are send and receive, and the Camera has a bit of Object detection written into it - fairly easy to do even with small CPU usage...

The bottom camera is used when flying under 25 feet, say flying indoors along with the IR sensors.

The camera on the bottom can "see" the floor since there's no GPS to help stabilize the Spark.

I have read that the bottom camera also takes a picture after lift off and retains the image in memory so that if a RTH is initiated, the Spark can compare the image in memory and the landing site to help with the "Precision Landing".

It's been stated that most pilots will launch straight up and hover for a few seconds at about 25 feet while the Spark snaps a picture.

 

[LKLM-Media]

The bottom camera is used when flying under 25 feet, say flying indoors along with the IR sensors.

The camera on the bottom can "see" the floor since there's no GPS to help stabilize the Spark.

I have read that the bottom camera also takes a picture after lift off and retains the image in memory so that if a RTH is initiated, the Spark can compare the image in memory and the landing site to help with the "Precision Landing".

It's been stated that most pilots will launch straight up and hover for a few seconds at about 25 feet while the Spark snaps a picture.

Now that you mention it I think I might have brushed past a comment regarding the Comparitive Image being taken, When landing I often take control anyway - sometimes landing on a platform I have on the back of my Ute/Truck (which I can't take off from as it screws the compass.)
I did some work a few years ago when I started working on entering the Outback Challenge here in Australia - some of the work was going to incorporate image processing done by Andrew Tridgell and his team at CanberraUAV - they were doing area passes and processing hundreds of images to look for "Outback Joe" - that was from height though, and using a 1st Gen Raspberry Pi. I imagine the Cortex or whatever is in the Spark is significantly bigger and better.