Calculating sampling frequency of a 5 min time stamp of GHI measurements

hello, can anyone tell/help me what the sampling frequency is of a 5 min time stamp of a series of GHI measurements? I have data of 1 month.

[Guillaume]

No idea what you're asking.

• What is a 5 minute timestamp? A timestamp doesn't usually have a duration. It's just a point in time
• What do you have as data? Provide an example either as a file or as valid matlab syntax, NOT as a screenshot.
• What do you want out of that?

[Caroline]

I mean I have data points every 5min and I need to know how to calculate the sampling frequency of a time series that records data every 5 min.

[Caroline]

I am using the low pass filter(bilinear transform to filter the data, but I need to know what the sampling frequency is..or how to calculate it.

Answer 1

[Star Strider]

If the sampling interval is 5 minutes (minutes per sample), the sampling frequency is the inverse of that, or 1/5=0.2 samples/minute.

The Nyquist frequency (the highest uniquely resolvable frequency in a sampled signal) would be half of that, or 0.1 samples/minute.

[Caroline]

So I can say that Fs = 0.2Hz?

[Star Strider]

No.

The sampling frequency for your data is 0.2 samples/minute.

It makes no sense to convert to Hz here. However if you want to, the conversion would be:

samples/second = samples/minute * minutes/second

or:

samples/second = 1/5 * 1/60 = 0.00333333333333333

so 0.2 samples/minute is the same as 0.00333333333333333 Hz, and the Nyquist frequency would be 0.00166666666666667 Hz.

It would be easiest to leave everything in terms of samples/minute and design your filter that way. The mathematics of discrete filter design are irrelevant to the time units used.

To design your lowpass filter in terms of samples/minute, if you want a cutoff frequency of samples/day, the calculation would be:

fco = samples/minute * minutes/hour * hours/day
fco = 1/5 * 60 * 24
fco = 288

You simply need to think through the unit conversions. This is much more convenient than adding the additional conversion to Hz, in addition to which the values themselves could become unwieldy and prone to precision issues as well.

[Caroline]

thank you very much now I get your answer. I asked this question because the filtered graph did not make sense at all.

[Caroline]

%1.filter function in Matlab H(z)=b[]z^-1/a[]z^-1 -> bilinear transform

x = time series data

K=1;

Ts= 5

Fs=1/Ts;

h=6/(2*pi);

B=(2*Fs*h);

b= [K/(1+B) K/(1+B)];

a= [1 (1-B)/(1+B)];

Ffilter1=filter(b,a,x);

So if I design the filter like this then Fs in that case = 0.2?

[Star Strider]

As always, my pleasure!

EDIT —

‘So if I design the filter like this then Fs in that case = 0.2?’

That is how I would do it. Keep all the current sampling times and time units unchanged.

Answer 2

[Guillaume]

I need to know how to calculate the sampling frequency of [whatever] that records data every 5 min.

Huh? If you record with a period of 5 min = 300 s, the sampling frequency is the inverse of that so, 1/300 Hz. Not sure what that has to do with matlab.