[greggold]

Why do this? Flight log plots let you see details of the radio signal performance that were previously unknown to the pilot. This allows adjustments or repairs of the rc system to be made on your valuable model before a signal failure and crash occurs. Simply put, flight logs are a major advance in protecting valuable models.

I have previously posted some flight log plots that show some things they can do. I will provide details of how I use google sheets to make a plot. If you can do a google search, you have google sheets on your pc or mac. If you can write flight logs on your tx, you are ready to go. Once you have made your first plot (it takes around 30 - 40 minutes) it will only take you <5 minutes for any subsequent plot.

I will give the instructions for a Radiomaster Tx16s tx, using a matek rx that has a single antenna. Different rxs and different gear brands require slight changes in the details but basically use the same instructions.

The instructions are broken down into 4 steps, with some minor preparation before step 1 and a few tweaks of the plot at the end.

Preparation

You need a special function to record a log on your tx sound card. I assign a switch to switch on write logs before launch and switch it off after landing. This creates a file in the Logs folder on the tx sound card, that is a .csv file. My logs always use a 0.5 second measurement interval, ie, 2 measurements per second. This file is moved to my laptop using a USB-C cable.

If you open google search on your computer, at top right there is a button called google apps, click it and find google sheets, then click sheets to open it. In sheets, click File/Open/Upload/Browse; this allows you to find the folder where you stored the .csv file, select the file and open it. If your flight lasted 5 minutes, you will see there are 600 rows in your spreadsheet, and you will see in the column headers the kinds of data in the sheet.

Step 1 Insert Columns (create the insert column macro)
These instructions will insert a column to the right of the following columns:
B,C,E,F,I,M,S
These cols have the headers: Time, 1RSS, RQly, RSNR, TPWR, RxBt, Thr
Click cell A1, then click Extensions/Macros/record. A small macro  box is opened while the macro runs, check it has absolute references selected.
Move the cursor to the header cell C, and find the down arrow near the right edge  of the cell, click it and select insert 1 column left. Note the 1RSS col has changed from col C to col D, and see how all the other columns have changed their header letter. Repeat this for cols E,H,J, N, S, and Z
Click the save button in the macro box, name the macro InsrtCol, and give it number 1. (Notice the spelling of InsrtCol without the e).

It is now essential to test this macro, as follows: find the undo button next to the magnifying glass in the top toolbar, and click it 6 times. Check all the new cols have gone, except the one beside Time col, and click undo once more. The spreadsheet is now the same as when loaded. Now run macro 1.
The method to run macro 1 on my mac is to press shift/option/command with one hand and press 1 with the other.
NB The first time a macro runs, you see a warning that macros can be dangerous, and google wants to check you trust the person who wrote it. You click allow to run - seeing as you wrote it yourself -  then you try to run it again and it works.
When the macro stops running, check cols Time, 1RSS, RQly, RSNR, TPWR, RxBt, and Thr now all have a blank col on their R side. If not, restore the spreadsheet (or delete all rows and reload the csv file), and start again.

Soon I will post Steps 2, 3 and 4.

Step 2  Insert Headers for the new columns
Step 3  Fill the new cols, eg by dividing the col to its left by 2, or other manipulations
Step 4  Insert the plot

Once you have the step 1 macro running, and tested, you are over the biggest hurdle and its downhill the rest of the way. If something is not working be sure to ask a club member to help.

[greggold]

Step 2 Insert Headers for the new columns

Check your sheet, and ensure that the following cols are blank:
C,E,H,J,N,S,Z
(If not, its best to repeat step one which will be quicker the second time around)
These are the headers to be added in row 1:
col C (no header - blank), col E RSS/2, col H RQly/2, col J RSNR*2, col N  TPWR/2, col S  RxBt*7, col Z  Thr/70

Click A1, then click Extensions/Macros/Record macro
Check the macro box has absolute references on
Click cell E1  (note we leave C1 blank)
Type 1RSS/2 then press return
Click cell H1 and type RQly/2, then press return
Continue on the list of blank columns till col Z header is done

Click Save in the Macro Box, name the macro: Header, and make it number 2
Click save in the macro box.
Test macro 2 as follows:
click the UNDO key 5 times. Check all the new headers have gone, except one. Click UNDO once more. Check the spreaadsheet is returned to the end of step 1, ie with the above 7 blank columns.
Click A1, then run Macro 2. Check cols E,H,J,N,S,Z have the correct headers

[caffeine]

This might be useful to demonstrate using screen recording software, so you can talk about what you're doing as you go.

I can recommend ActivePresenter for this. There's a free version which creates video screen captures without watermarks

https://atomisystems.com/download/

[greggold]

Step 3

Fill the new cols C,E,H,J,N,S,Z, for example by dividing the col to its left by 2, or other manipulations. A new Macro 3 is called Divide, however sometimes the action is add, sometimes its multiply. Do not type in the bracketed comments when the macro is being recorded.
Click A1, then click Extensions/Macros/Record macro
Check the macro box has absolute references on
Click cell C2, type 0, enter (its a zero, not a capital o)
(this brings the cell selected to C3)
In C3 type = then click on cell C2, then type +0.5 then press enter
press command and click enter on a mac (Command-enter does the autofill of the entire column), on a pc I believe its press control - enter
Click E2
type = then click on D2, then type /2 enter
press command-enter (or control-enter on a pc)
Click H2
type = then click on G2, then type /2 enter
press command-enter (or control-enter on a pc)
Click J2
type = then click on I2, then type *2 enter
press command-enter (or control-enter on a pc)
Click N2
type = then click on M2, then type /2 enter
press command-enter (or control-enter on a pc)
Click S2
type = then click on R2, then type *7 enter
press command-enter (or control-enter on a pc)
Click Z2
type = then click on Y2, then type /70 enter
press command-enter (or control-enter on a pc)

Click SAVE in the macro box, name the macro divide, and assign number 3
click save the macro

As usual, its important to test the macro. Click the UNDO key repeatedly, watching the column content disappear until in col C you see the zero disappear

Run macro 3

If everything is working, you are now ready for step 4 where the plot finally appears.

[greggold]

Thanks Liam I am interested to look at Active Presenter when I have finished step 4. However I am a bit old-school, and find text instructions are often clearer, more efficient and more accessible than say a 15 minute youtube video. But I agree I should look at Active Presenter. I would appreciate any comments on the above steps, do you use a pc or mac? If a pc, does the autofill function work by pressing control-enter? Have you used google sheets on a pc?

Greg

[greggold]

Step 4
Insert the plot

The 7 columns to be selected are C,E,H,J,N,S,Z
Select column C by clicking the letter C in the top of the spreadsheet; the col will change color or tint when selected. Then select the remaining cols by holding down the command key on a mac (control key on a pc) and click on cols E,H,J,N,S,Z
If you now release the command/control key, you can see the 7 cols remain selected if you lightly scroll sideways.

Click insert/chart

Click save in the small Macro box
name the macro insrtChart
click save

I will now detail the tweaks I apply to each chart using the chart editor. When I am better at editing java scripts (the App Script) I think all the tweaks can be included in macro 4, but I have not yet been able to accomplish this in a simple reliable manner. In the future all the 4 steps can be done in 1 macro, but for starters I think it is helpful and easier to break down the task into 4 small components with 4 macros.

PLOT TWEAKS
1. Click the top right corner of the chart to activate a button, then click edit chart to open the Chart Editor. It has 2 main menus, Setup and Customise.
In SETUP find the entry column chart, and change it to scatter chart which you find by scrolling down.
Then click CUSTOMIZE, and Click on SERIES, and under Apply to all Series you find format/pointsize is set to 7; change it to 2
Click on SERIES, and a drop down menu allows you to select a single series to edit. I change thr/70 to a pale grey color, and RxBt to a light purple, as these 2 are not the prime indicators of a good signal.

Scroll down to VERTICAL AXIS, change the axis min to -50 and max to 52, which allows the RQly series to be fully displayed.

Click GRIDLINES AND TICKS, and the vertical axis settings are displayed. Click minor gridlines box, and set minor count to 4.
Change the vertical axis settings to horizontal axis, by clicking the down arrow next to vertical axis. In the box next to the step box, set the step value to 50. Click minor gridlines box, and set minor count to 4.

[greggold]

Additional points on making a plot.

When you have finished tweaking a plot, including giving it a title and naming the horizontal axis time, download the plot as a pdf. When making the plot title, copy and paste the name of your spreadsheet as it exactly identifies the data file.

Make a copy of your spreadsheet, and rename either the original or the copy as the logger or the log-plotter spreadsheet. The copy has all the macros. Then delete all the rows in the logger file, by selecting the first row, and pressing shift-row header on the last row, and use the edit/delete function. Note that the macros are not deleted in doing this.

Then shut your first plot down, and click file/open/upload/browse to open your second log file. With col A visible, scroll down to the last row and click shift-row header so you select all the rows in this spreadsheet. Press command-C (control-C on a pc) to copy the rows, then you select cell A1 in the logger file. Paste the rows by pressing command-V. A box appears asking if you want to paste. Click enter, then you have to click command-V a second time, and wait around 5 seconds for your new spreadsheet to appear. Then click on A1. Then run the 4 macros, each taking only a second or several seconds to finish.

[caffeine]

Based on your instructions here, I have started making a web app to automate the process;

   

[greggold]

Based on your instructions here, I have started making a web app to automate the process;

Its looking good!

Your line plot works well, my plots are a scatter plot and which is the better choice of plot type is unclear. Does Active Presenter which I expect is what you are using allow scatter plots?
This is a real test for Active Presenter to put all the data series (the RSS, RSNR etc) on one plot. As you see in my plots, I had to manipulate every data item (eg divide it by 2) to achieve this. I considered having a second y-axis, but my experience with plots is a single y-axis is clearer.

In relation to step/macro 4, I have added some java code to macro 4 in my Google sheets Apps Script so the plot comes out in a near finished form. However this code may not work in Active Presenter which likely has its own script language. I have not posted how to open Apps Script and add the lines of code, but I can do this if you or anyone wants to try it. Now I can go from loading the csv file to finished plot in a few minutes and I will be logging/plotting most flights on all my planes from now on to see what this achieves.

Here is one question I will be able to answer soon: for our field, is it better for elrs flyers to have a dynamic power range of 10 -100mw, or from 25-100 mw? The test is in the RQly, and its looking like the RQly remains higher throughout the flight with the 25 minimum. However I need to confirm this with a better analysis. I do not have any type of glitch or signal failure occurring in any of my elrs planes, so maybe the range does not matter, but I think better rx quality is always worth having. I still get the occasional crash but now I can show its not due to signal failure, so I wonder what that means....

If the analysis shows 25 minimum is better, I will try the sunjun technique I posted to set the minimum dyn power on a Tx16S to 25mw. I have tried to ask Discord to comment on sunjuns technique, but there was no response. However his method seems very simple and reversible, so I am prepared to give it a go.

[caffeine]

I'm not using Active Presenter for this, that's screen recording software. I'm writing this using node, react, typescript and google charts.

I like the idea of scaling the plots to be in the same range. Perhaps some mechanism to automatically scale everything to range 0-100 would be suitable.

For the idea of setting a base TX power of 10 mW, are you concerned that the dynamically changing power may not respond quickly enough? If not, then what is your motivation for reducing the occurrence of TX power increases?

I'll modify my web app to be a bit smarter about loading and parsing the CSV file, then I'll put it on my webserver and you can try it out if you'd like.,

[greggold]

Interesting points here.

re: using node, react, typescript and google charts

I do not not know what node, react, typescript are. What is the application they are located in?
A big strength of google sheets is that it both a powerful spreadsheet and a powerful graphics app, and one program does it all. LibreOffice Calc is not as good with the graphics. Excel would likely be suitable if you have a Microsoft Office subscription. In my mac there is no separate google charts app listed in the google apps menu, and I expect you meant google charts within the google sheets app, is that correct? Once you click insert chart in google sheets, a plot appears in your spreadsheet and by clicking top right on the plot you open a Chart Editor.

re: I like the idea of scaling the plots to be in the same range.
I do not like this so much, as I want the number as shown in telemetry and you can easily get it with a simple multiplication or division by 2 for the key signal strength indicators (RSS, RQly, RSNR and TPWR) using the above macros. I always want to plot RxBt and a 0-100 range will have a problem on the Tx16s as RSNR and TPWR will overlap each other, and RxBt is also in the 0-10 zone so its crowded (RQly and Rxbt would remain at near 100%, and RSSI and RSNR would be overlapping at around 70%). Having a max value of 52 for the y axis gives a clearer picture of each series. But I am sure people will have individual preferences on the plot customization.

re: For the idea of setting a base TX power of 10 mW, are you concerned that the dynamically changing power may not respond quickly enough?

Its a good question and the Discard group raised the same issue. The reason for a 25mw base dyn power is if the overall signal quality is better with 25mw, as judged by RQly (most important), RSNR (next most important), and RSSI, surely you would choose to have it, especially if choosing 25mw over 10mw has no downside. I think there is no downside except I have no certainty that sunjuns method to set it does work. I have the impression 25w is better, and I will do an analysis to check this. You can never have too much rc signal quality in a valuable model.

Having said that, the dynamic power function in my experience over many flights over many months is that it does respond quickly enough, and the plots always show this. At the risk of being boring/repetitive, ELRS is so much better than any rc system I previously used, with dynamic power rated as the best feature.

[caffeine]

node, react, typescript are web based technologies. There is no application. Everything runs in a web browser.

I have put an early version up at https://autographer.liamohagan.com.au/

Drag and drop a log file onto the section where it says 'Drag and drop here', or copy and paste the actual file into the window, or click on the 'Browse' link.

Once a log file is selected, it will be graphed.

   

If you scroll down, there are additional columns that can be added to the graph, click the checkbox and the graph should redraw.

At the moment the check boxes don't update, I'm working on that, and the scaling doesn't do anything yet. I think I will add offset as well.

I'll make it so that the selected columns, scaling and offset per signal will be stored in your web browser settings, so it will remember the settings for each run.

I will also add a button to export the graph, add titles etc.

If you want to look at anotehr log, click the little x to remove the previous log, and upload another one.

[greggold]

node, react, typescript are web based technologies. There is no application. Everything runs in a web browser.

I have put an early version up at https://autographer.liamohagan.com.au/

Drag and drop a log file onto the section where it says 'Drag and drop here', or copy and paste the actual file into the window, or click on the 'Browse' link.

Once a log file is selected, it will be graphed.



If you scroll down, there are additional columns that can be added to the graph, click the checkbox and the graph should redraw.

At the moment the check boxes don't update, I'm working on that, and the scaling doesn't do anything yet. I think I will add offset as well.

I'll make it so that the selected columns, scaling and offset per signal will be stored in your web browser settings, so it will remember the settings for each run.

I will also add a button to export the graph, add titles etc.

If you want to look at anotehr log, click the little x to remove the previous log, and upload another one.

It is really cool ... I attach my first Autographer plot. Its nice you can select the number of cols and rows to plot, and each data series adopts the name provided in the csv file from your tx. Can I ask how you are programming or scripting this; and what editor do you use to code or script?

[caffeine]

I'm using a tool called create-react-app to make the react/node/typescript framework. Then it's a matter of editing the typescript files to generate the page.

For the editing I use visual studio, but any text editor will do it.

Node/React has a plugin architecture (example https://www.npmjs.com/package/react-google-charts) , and I'm using material UI for the controls, filepond for file uploads, papaparse to read the csv and react google charts to display the results.

There's a lot more that can be done with interacting with the charts, I'm reading up on how to zoom in on interesting sections. An example here: https://codesandbox.io/s/github/RakanNim...nge-filter

[greggold]

I really like what you have done with Autographer. The way I suggest you could proceed is to have a separate spreadsheet do the preparation of the write log csv file, called logfile.csv
(a) add columns and manipulate any data item, eg, the time column could be changed into seconds, (the way step/macro 1 and 3 does), so a 5 min flight will have an x-axis of 0-300
(b) add or edit the headers

Then the logcsv.csv file be saved as logfileED.csv.

This separates the preliminary spreadsheet manipulations and spreadsheet functions from your Autographer project, and allows Autographer to focus on the plot. You can use any spreadsheet for (a) and (b), my reason for selecting google sheets was that it is one of the few spreadsheets with a great graphics capability. The preliminary steps (before starting to plot) can readily be automated as macros, a function which all spreadsheets have, AFAIK.

With the logfileED.csv file loaded, you can then immediately visualise your plots, and Autographer already lets you easily select "data series" (eg a data serie is the RSSI column of the spreadsheet), pair any serie with another serie or series, and presents a key of what color represents each serie. I think already you can select an x-axis range of say from 100 to 150 seconds which you might want to do if say you are analysing a crash that occurred at 150 sec into the flight.
My suggestions going forward would be to allow changing or specifying the color of a selected data series, and the line thickness. If a scatter plot type can be an option, together with a line plot option, this would be a nice feature but not essential.

Plots are really complex. You also could add capability to label the x-axis and the plot title, add gridlines, put numbers on the gridlines (eg, the x-axis has a gridline every 10 sec, with gridlines 0, 50sec,100, 150.. labelled) and to save the plot as a pdf file. However maybe these are not essential to quickly see what the plot is saying.

[caffeine]

Yes, all of those things are possible.

I also thought some analysis might be interesting. For example, what was the flight duration, total battery consumption, etc

On a somewhat related note, I found today that sudden throttle cuts sometimes correlated with an increase in tx power, just as you described recently. Interesting stuff.

I'll see what the plot of tx power vs throttle channel looks like

[caffeine]

I've updated the tool to allow selection (and deselection) of which items to plot, and also added a per-series scaling factor.

Next I will add a per-series offset mechanism, so you can move a plot up and down the graph without scaling.

   

[caffeine]

Time is now in seconds (or milliseconds for very short flights ) since the start of the log, and you can now scale and offset a trace up or down the graph.

   

[greggold]

Autographer (AG) and Google Sheets (GS)

Liam has shown AG is now a practical tool for viewing flight log files, and we can compare Liam's AG plot method to my google sheets (GS) method using an interesting flight log from a 1m span 3D plane. Both AG and GS are works in progress and have recent improvements not reflected here.

A big strength of AG is that it reads your log csv file and goes directly to the plot. With GS you have to move the csv file to a GS logger file, and run macros.

The following plot analyses are a bit long-winded but I would urge people to read them as they show aspects of rc signals that have not been previously revealed.

The GS file is a pdf file named Super Extra. It shows the transmitter power (orange, TPWR/2) at around 80 seconds has jumps from the baseline of 10 mw (orange line) to 25, 50 and 100 mw . The AG also shows the jumps at 80 sec. However the GS has more detail: it shows the jump first went at 79 sec to 100mw, then to 50mw at 79.5s, then to 25 at 80s. (AG could zoom in on the 60 - 90 seconds interval to show the detail). Both GS and AG show TPWR then returns to base line 10.

AG will soon have throttle power included. GS shows here that the TPWR jump to 100 occurred BEFORE full throttle (grey dots) was applied at 82 sec (for the vertical climb of the blender manoeuvre). This means the throttle increase did not trigger this jump. At 79 sec when the jump started, the RSNR*2 (radio signal to noise) was within normal limits at 5. At 79 sec is a sudden drop in RQly (receiver quality aka valid frame rate, the percent of data packets that are valid) to 78%. (AG has this RQly drop but its a bit unclear, however a zoom in on the 60 - 90 seconds interval would make it clear). We know from the elrs website that jumps in power are triggered by drops in RSNR, RQly, and RSSI, yet at 79 sec RSSI (on both antennas, RSS1 and RSS2) did not fall. A RQly of 78% is within limits and cannot cause signal loss, so here the decision to jump to 100 appears to be a precaution based on a sudden fall in RQly.

The blender is suspected of causing drops in receiver signal quality and RSNR and that appears to be the case here. The blender climb started at 82 sec and throttle was at 10 (almost full) on the plot till 91 sec when the top  of the blender was reached and throttle was cut to -15. In the wild spin and gyrations that followed throttle was increased to 15 (absolute full). The high throttle, and the gyrations look like they may be associated with the falls in RSNR. 

From 82 sec through to 225 sec both plots confirm there were big falls in RSNR. At 145 and 152 sec RSNR fell to zero, (below 1 values are abnormal) yet at these times the RQLY and RSSI were fairly nominal (within normal limits). At 165 sec there is a TPWR jump to 25 mw, that lasts at least 1 sec, it is associated with a drop in RSNR to 1 clearly shown on AG, but not clear on GS as the green dot is obscured by a grey one (once again, zooming on the time interval is simple and would make this clear).

From 230 to 330 seconds a big question is posed: why is this plot now totally different? The RSNR is now always 12 or 13, the RQLY is 100%, and RSSI is higher. The blender (and other wild manoeuvres) seems to have an effect, I see it most flights, and other aerobatic pilots have reported an effect,  but it remains mysterious.

With AG and GS methods to view logs, there is finally a chance to figure this out.

[caffeine]

Greg, could you attach the log file here? So that I may work with the same data to try to improve Autographer.

I see lots of TPWR increases correlated with sudden, dramatic drops in throttle signal. Perhaps there's lots of electrical noise when the motor is being throttled down.

[greggold]

Greg, could you attach the log file here? So that I may work with the same data to try to improve Autographer.

I see lots of TPWR increases correlated with sudden, dramatic drops in throttle signal. Perhaps there's lots of electrical noise when the motor is being throttled down.

here is the log file ...  oops  it says this file type is not allowed to be attached, hey its only a csv and 109 KB. Can you send me by whatsapp or pm an email address.

I saved the csv as an excel file and webpage is happy. Let me know if the file is unreadable.

[caffeine]

Converted it back to csv and it loads fine, thanks!

   

[greggold]

re "I see lots of TPWR increases correlated with sudden, dramatic drops in throttle signal. Perhaps there's lots of electrical noise when the motor is being throttled down."

Comparing 2 intervals, 75 - 230 sec and 230 sec -340 sec, there is higher throttle overall in 75-230s, with more ups and downs. The ups are higher, but apart from one down around 90s, the downs are not really different to 230-340 interval (ok the down at 160s is a little bigger). The standout thing is that 75-230 has lower RSNR, yet the signal indicators are not showing much. The signal has little change, so for the signal-to-noise ratio to fall, the noise must be more. I agree with you, I suspect the throttle, but I do not know whether its turn up, turn down, or average throttle that is responsible.

[caffeine]

I have made the graph take up more of the usable screen space, I have also added the filename and flight duration to the title.

   

Lastly, I have added the ability to zoom and pan around the graph to take a closer look at anything that's interesting. Mousewheel to zoom, click and drag to pan, and right click to reset.

   

[greggold]

Some really good log plots here, including some zoom in. And I am impressed you are getting 9 minutes of flight time. I have not seen your plane nor heard about it, I wondered if I could tell anything about it from the plots. Whether the plane has a 2S or 3S lipo: I find with 2S lipos and a matek rx the RSNR does not drop to zero, as it has done here on 5 occasions during the 9 minutes but it very quickly bounces back. RSNR falls to zero may be more common with 3S planes. However the lack of any change in any signal parameters with big throttle ups and a down makes me place a bet its a 2S. Its not a 3D plane, they have a very different throttle profile with frequent big changes, up or down, and I see changes in RSSI and RSNR when thr changes.

The plots are doing nothing for the theory that the throttle changes up or down cause tx power jumps. However the falls in RSNR appear a bit related to the jumps: since your signal remains normal (based on RSSI and RQly) while RSNR falls, the ratio of signal to noise may fall due to bursts of noise.

I see you have dynamic power on, and base value of TPWR is 10mw, so I rule out you fly with a RM boxer or a TX12 tx, and I think you must have a RM t16s.

Throughout the flight the RSSI and the RQly are nominal, and really good. On 2 occasions your tx saw the need to jump to 50 mw, and around 15 times to jump to 25 mw. I have had flights with such frequent jumps, and I have decided the jumps reflect the fact that PRCAC is a noisy radio environment, causing falls in RSNR. There are 2 viewpoints:
(a) the tx is working as per its design, no need to worry as RSNR is quickly restored after a jump
(b) if you set the base dyn power to 25mw, you would see a very different log, a near pristine log with basically nothing to see. I know this because I have a tx12 as well as a TX16s, and TX12 has base dyn power of 25mw. I am still considering whether to increase base dyn power on my Tx16s. I don't like RSNR values of zero, no matter how transient.

In another thread I discussed the views in Discord, and Sunjun's method to set the base dyn power to 25mw.

I suggest in plots instead of adding 10 to RSNR, you multiply it by 2. This is because a fall in RSNR of 10 (from 12 to 2) is a very different thing to a fall in RQly from 100 to 90; also the RSNR may be the most sensitive indicator of what is happening with the signal. The RQly and RSSI are nominal throughout while RSNR falls to zero on 5 occasions.

[caffeine]

Good detective work! It's a 3S 3200mAh LiPo in this particular plane, powering a 3526 motor with 830kV, attached to a 11x8 prop.

Here's an even better plot:

   

My Kmart foamie with an 850mAh 3S, nearly 15 minutes with plenty of battery left!

[greggold]

Its a really informative plot, and I like your RSNR method, RSNR=2*RSNR+20. Since 20 is a gridline, you can immediately see if RSNR drops to zero. Great work.

[caffeine]

I did another flight today, with very gentle throttle inputs, to see how long I could fly (over 16 minutes with 700mAh used of 850aH capacity), and to see if there was any correlation between throttle movement and TPWR changes (there was not)

   

[greggold]

Your 934.csv log is really interesting. Interval 150-200 sec: after a 12 sec full power burst, there is a sustained drop in RSNR with a drop in RSSI and RQly. However there is only a momentary power jump to 25 mw. This compares to around 620 sec where there is no change in throttle, but an unexplained drop in RSNR to zero, followed by a fall in RQLY, and a power jump to 100mw. Could you send me this csv file (maybe as whatsapp attachment), I want to zoom in etc. Do you think the drop in RSNR happened first and triggered the other changes including power jump?


The 934.csv log makes me a little nervous, why does RSNR keep falling to zero?
The values of RSNR are different to the 318.csv log, where only 1 zero value was recorded.

[caffeine]

I'll attach it later when I get to my PC.

The full power section was a vertical climb to altitude (in the 'danger zone' at PRCAC), so I suspect that was what precipitated the change in RSNR.