Drone Data Processing stuck at FLOAT

Hi everyone,

Looking for some guidance whilst trying to process our drone mapping data.
We are using the Reach RS2 as base and have attached a Mettatec X5 to our Phantom 4 Pro drone.
Based in NZ, using Emlid Studio 1.7

When processing with Emlid studio drone data default settings, it gets stuck a quarter of the way through at FLOAT 00:23:29
When turning off all satellites except GPS we get a 95% FIX solution.
Does anyone know why this is happening?
How can we get a 100% FIX solution, is this necessary to then geotag our drone mapping photos?


Hi Jemma,

Welcome to our community!

As per your screenshot, there are some weird values in the plot: 06 Jan 1980 observation start, the baseline is up to 12109.6 km. Most likely, they are related to the information registered in the RINEX header, so it’s worth looking closely at the files.

Could you please post your raw data logs or send them privately to support@emlid.com, so I can check what could cause it?

1 Like

Hi Kirill,

Thanks for your response, it is very strange.
Here is the raw data files:

We have had problems like this in the past also and # of events not lining up with # of photos from the drone

Jemma

Jemma,

Initially, I reproduced your issue: Emlid Studio stuck at the exact second. It made me think that the software couldn’t handle such a large amount of data, so I decided to thin out the base log to the 1 Hz update rate. For processing, I ticked check boxes of the Logs duration as shown in the screenshot below:


And it worked out. Emlid Studio processed the data successfully with 99.9% of FIX solutions.

I’ve attached both the resulting pos files
in the zip archive below, just in case. You can try to process with the same settings and let me know how it goes.

Jemma’s processed files.zip (4.4 MB)

I’ve found the invalid time mark in the raw data log:


However, it was processed by Emlid Studio.

At this point, I suppose you can contact Mettatec regarding this issue since their receiver is installed on the drone.

Thank you so much Krill that is extremely helpful!!

How did you thin out the base log it 1 Hz update rate?
And how/what software do you use to look at the log data to find the invalid time marks?

Jemma

I think you can open via notepad/notepad++ to read that file

1 Like

You can do this in the converter tool of Emlid Studio.

As correctly mentioned below, you can do this in any text editor like Notepad. But I used Visual Studio Code.

1 Like

For our next project we are having similar issues, getting stuck halfway or giving inconsistent fix/float results
(I had to combine a UBX file for the base Reach RS2, hence the first screenshot having a lot of single results) but other areas are giving inconsistent results



And another one just pausing half way for no apparent reason :((

Jemma,

It looks like you’re using the RTKCONV to convert UBX to RINEX, which generates this odd time stamp of 06.01.1980. See the same issue in this thread:

Other than that, please share with me your datasets again, so I can reproduce the issue myself.

Here is the dataset for the Matakana project;
there are 2 different properties - we are having issues with processing both

The we flew another 3 flight plans for the original project that we are also still having issues with

Jemma,

I’ll split my response into 3 sections as per each shared dataset.

Clayden Road

This dataset was processed smoothly on default settings of Emlid Studio and using navigation file from the base. For some reason, the plot shows the odd 06.01.1980 date and weird event with zero coordinates, but the events.pos file doesn’t have it.

The event.pos file is below.
X5_logs_231201_025932_events.pos (209.8 KB)

Matakana Valley

Emlid Studio stuck at the same moment, so I’ve decided to split the processing in two parts. The first lasts until 01:15:00, while the second starts at 01:16:00:


The reason of such distinction is that there’s a gap in a raw data at this moment, that could cause such an issue:

I’d note that all raw data logs from the drone have such gaps. Particularly, a long gap of approximately 10 minutes in this log:

Again, I think it’s a question for the drone or receiver manufacturer.

Events.pos files are below.
X5_logs_231201_004912_1part_events.pos (50.6 KB)
X5_logs_231201_004912_events_2part.pos (103.0 KB)

TW Day 02 Processing

Here’s something wrong with the base file: it only lasts for 4 minutes and contains too few satellites:


If there are no other base logs, a highrate log from the WARK00NZL0 station of AUSCORS could help in theory:

However, I haven’t managed to find any logs from this station in the public data.

Probably, you can request the log for the required time period by emailing to AUSCORS.

1 Like

Hi Kirill,

Thank you that’s very helpful again.
Apologies the TW Day 02 had the wrong reach logs in it so these are the right ones now:

I’ll give it a go myself.

For future projects, can we assume that if during Emlid processing it gets stuck at a certain point we are best to split the data then and exclude the gaps?
I will get in touch with Mettatec about potential issue with the X5 creating gaps and weird starting dates (1980).

This way everything is processed fine without any odd time stamps and with no freezes on default settings.

You can find both pos files attached.
X5_logs_231130_220931_events.pos (198.0 KB)
X5_logs_231130_220931.pos (9.6 MB)

Thank you so much Kirill, you’re a life saver!

Next question, (I’ve contacted Metatec as well), but I was wondering if you could help find these missed events and therefore photos I need to remove in order to geotag them in Emlid.
If you could also explain the workflow for finding missed events in Visual Studio Code then I would be able to do it myself in future if this issue keeps occurring.

This flight TW Day 02 requires 1,431 photos to geotag in Emlid however we have 1,450 so I am unsure which ones to remove in order to tag.

Thanks heaps

Drone images for these ^ Day 02 attached

Jemma,

You just need to do string searching in the VS Code: Ctrl+F, input not valid or time mark. Just as shown on one of my screenshots above in this thread.

As for searching for photos to remove, this can be a really complicated task considering the large number of photos. Usually, you can see some gaps in the drone’s trajectory on the plot, which can indirectly indicate that there are missing events. But such gaps aren’t visible in your dataset, and you also have to compare the timestamps of the photos with the processed events to be as accurate as possible. So I’d still recommend you reconsidering the hardware setup to avoid such hard processes in your workflow.

Thanks for information