Matlab does plot datetime data (only hours) on wrong day

Hello,

I have data from two sensors that I want to plot in the same window to compare differences in my X value.

I have them both stored in tables and use the datetime "HH:mm:ss.SS" format.

For example:

sensor1.Time = ['12:41:50.00'; '12:41:51.00'; ...]

sensor1.x = [0.01; 0.02; 0.02; ...]

and

sensor2.Time = ['12:42:05.46'; '12:42:05.49'; ...]

sensor2.x = [0.01; 0.02; 0.01; ...]

When I plot the data using

plot(sensor1.Time, sensor1.x) it plots the data in a window and adds todays date.

However, when I plot the other table it adds a random date (here: 15.07.2022).

If I plot both using hold on; the data then looks like this which is not very helpful:

How can I fix this date error and have the data aligned on one day?

Best Regards,

Renan

Answer 1

[Star Strider]

Without having the data to experiment with, my only suggestion is to plot them after converting them to duration arrays (that itself is not as straightforward as I would like it to be, especially since duration arrays are limited to one day in length) —

Time1 = datetime(2022,7,15) + minutes(0:1439).'

Time1 = 1440×1 datetime array
   15-Jul-2022 00:00:00
   15-Jul-2022 00:01:00
   15-Jul-2022 00:02:00
   15-Jul-2022 00:03:00
   15-Jul-2022 00:04:00
   15-Jul-2022 00:05:00
   15-Jul-2022 00:06:00
   15-Jul-2022 00:07:00
   15-Jul-2022 00:08:00
   15-Jul-2022 00:09:00
   15-Jul-2022 00:10:00
   15-Jul-2022 00:11:00
   15-Jul-2022 00:12:00
   15-Jul-2022 00:13:00
   15-Jul-2022 00:14:00
   15-Jul-2022 00:15:00
   15-Jul-2022 00:16:00
   15-Jul-2022 00:17:00
   15-Jul-2022 00:18:00
   15-Jul-2022 00:19:00
   15-Jul-2022 00:20:00
   15-Jul-2022 00:21:00
   15-Jul-2022 00:22:00
   15-Jul-2022 00:23:00
   15-Jul-2022 00:24:00
   15-Jul-2022 00:25:00
   15-Jul-2022 00:26:00
   15-Jul-2022 00:27:00
   15-Jul-2022 00:28:00
   15-Jul-2022 00:29:00

Data1 = sin(2*pi*(0:numel(Time1)-1)/720);

Time2 = datetime(2022,07,18) + minutes(0:1439).'

Time2 = 1440×1 datetime array
   18-Jul-2022 00:00:00
   18-Jul-2022 00:01:00
   18-Jul-2022 00:02:00
   18-Jul-2022 00:03:00
   18-Jul-2022 00:04:00
   18-Jul-2022 00:05:00
   18-Jul-2022 00:06:00
   18-Jul-2022 00:07:00
   18-Jul-2022 00:08:00
   18-Jul-2022 00:09:00
   18-Jul-2022 00:10:00
   18-Jul-2022 00:11:00
   18-Jul-2022 00:12:00
   18-Jul-2022 00:13:00
   18-Jul-2022 00:14:00
   18-Jul-2022 00:15:00
   18-Jul-2022 00:16:00
   18-Jul-2022 00:17:00
   18-Jul-2022 00:18:00
   18-Jul-2022 00:19:00
   18-Jul-2022 00:20:00
   18-Jul-2022 00:21:00
   18-Jul-2022 00:22:00
   18-Jul-2022 00:23:00
   18-Jul-2022 00:24:00
   18-Jul-2022 00:25:00
   18-Jul-2022 00:26:00
   18-Jul-2022 00:27:00
   18-Jul-2022 00:28:00
   18-Jul-2022 00:29:00

Data2 = cos(2*pi*(0:numel(Time1)-1)/720);

figure
plot(Time1, Data1)

figure
plot(Time2, Data2)

figure
plot(Time1,Data1)
hold on
plot(Time2, Data2)
hold off

Time1.Format = 'HH:mm:ss';
Dur1 = duration(string(Time1));                     % Convert To 'duration' Array
Time2.Format = 'HH:mm:ss';
Dur2 = duration(string(Time2));                     % Convert To 'duration' Array

figure
plot(Dur1, Data1)
hold on
plot(Dur2, Data2)
hold off

Your data will likely work for a duration array, however if they go for longer than one calendar day, it might be necessary to convert the time values only to datenum vectors and plot the data against that instead.

.

[the cyclist]

Great stuff, but I believe you are mistaken (or I misunderstood) when you state "duration arrays are limited to one day in length".

dur = datetime("15-Jul-2022") - datetime("13-Jul-2022")

dur = duration
   48:00:00

[Renan Deuter]

Hey,

thank you very much. This did work great. Thanks so much for this full response! Luckily my data is shorter than one day so this fix was fast-forward.

Best Regards,

Renan

[Star Strider]

Thank you!

Run my code with more than 1440 minutes (for example 0:2879). The duration plots wrap at one day, only displaying two cycles each instead of the four I’d expect, with a straight line returning to the beginning of each. (I already did that experiment, the reason I added that and my datenum recommendation to my Answer. I believe datenum would work correctly for data spanning more than one calendar day, however I did not actually do that part of the experiment yet. That might require datevec and etime with an additional conversion to minutes, depending on the precison of the original time vector.)

.

[Star Strider]

@Renan Deuter — As always, my pleasure!

[Steven Lord]

@Star Strider The behavior you described is correct, but it's not due to any issue with duration. It's due to the fact that the datetime data covers multiple days.

midnight = datetime('today')

midnight = datetime
   17-Jul-2022

dt = midnight + minutes([-5, 5])

dt = 1×2 datetime array
   16-Jul-2022 23:55:00   17-Jul-2022 00:05:00

dt.Format = 'HH:mm:ss'

dt = 1×2 datetime array
   23:55:00   00:05:00

If you extracted the duration data from the datetime array with timeofday, you could check for any large jumps. This would indicate crossing midnight and could be resolved by adding 24 hours.

tod = timeofday(dt)

tod = 1×2 duration array
   23:55:00   00:05:00

d = [0, diff(tod)]

d = 1×2 duration array
    00:00:00   -23:50:00

tod(d < 0) = tod(d < 0) + hours(24)

tod = 1×2 duration array
   23:55:00   24:05:00

check = midnight - days(1) + tod

check = 1×2 datetime array
   16-Jul-2022 23:55:00   17-Jul-2022 00:05:00

The check array looks like it matches the dt array.

Answer 2

[the cyclist]

The most fundamental fix to your issue is to explicitly use the modern datetime format for your times. Then, downstream functions that use them as inputs can be manipulated to control the display output.

Also, rather than "MATLAB randomly does something for some reason", one can generally find the exact behavior in the documentation for each function.

Answer 3

[Renan Deuter]

Another way that worked and might actually even be easier is this:

1) Create an new variable with a custom date

day = repmat(datetime(2022, 6, 30), length(sensor1.x), 1)

2) Merge two variables into one datetime array following this guide

https://de.mathworks.com/help/matlab/matlab_prog/combine-date-and-time-from-separate-variables.html

-> Basically we reformat both columns so both can fit hours and a date.

sensor1.Time.Format = 'dd.MM.uuuu HH:mm:ss.SS';

day.Format = 'dd.MM.uuuu HH:mm:ss.SS';

This sets our day variable to: 30.06.2022 00:00:00.00

and the Time column to: 15.07.2022 14:04:32.23

The we copy the time to the day variable

sensor1.Time = day + timeofday(sensor1.Time);