LabVIEW
Splitting and averaging of arrays
Solved!
Get the waveform components, average data by 100 and change dT by x100.
If you stay with waveforms instead of Dynamic data, you should simply have an array of waveforms to splitting them should be easy enough.
/Y
You start with a Waveform, and want to replace it with a Waveform that results from averaging over 100 points, resulting in a Waveform 1/100th the size (or at least that's what I assume you are doing). To have this work seamlessly, you'd want the size of the Waveform Y data chunk to be a multiple of 100, such as 1000 points. Then you'd simply peal off 100 points at a time, average them, and build a new waveform. There is no need for point-by-point averaging, since you (a) need to get the Waveform before you can process it, (b) this requires that all its points be available, so (c) there's no problem in "resizing" the array.
The processing is simple. Assume your waveform has size 100*m. Reshape the 1D Array into a 2D m-by-100 2D Array, pass this through a For Loop with a "Mean" function inside to convert it to an Array of size m consisting of the average of each sub-array of size 100. Use this to build a new Waveform, also (as Yamaeda noted) multiplying dt by 100. This example, based on a 1000 samples of a generated Waveform of N(0, 1) random numbers gives an Waveform where the resulting Waveform is now N(0, 0.1), as expected.
Bob Schor
@crossrulz wrote:
@KRICLA wrote:
Any suggestions on how to solve this issue?
Use a FOR loop around the Mean PtByPt.
Only if you use a parallel FOR loop with N parallel instances. Else all iterations will use the same instance of mean ptbypt, mixing all channels.
@KRICLA wrote:
-I want to compress this signal by a factor of 100, i.e. average all the samples,while preserving the waveform timestamp.
Well, each sample has it's own time, so you should probably calculate the timestamp of the central element instead of keeping the one of the first element.
@Bob_Schor wrote:
You start with a Waveform, and want to replace it with a Waveform that results from averaging over 100 points, resulting in a Waveform 1/100th the size (or at least that's what I assume you are doing). To have this work seamlessly, you'd want the size of the Waveform Y data chunk to be a multiple of 100, such as 1000 points. Then you'd simply peal off 100 points at a time, average them, and build a new waveform. There is no need for point-by-point averaging, since you (a) need to get the Waveform before you can process it, (b) this requires that all its points be available, so (c) there's no problem in "resizing" the array.
The processing is simple. Assume your waveform has size 100*m. Reshape the 1D Array into a 2D m-by-100 2D Array, pass this through a For Loop with a "Mean" function inside to convert it to an Array of size m consisting of the average of each sub-array of size 100. Use this to build a new Waveform, also (as Yamaeda noted) multiplying dt by 100. This example, based on a 1000 samples of a generated Waveform of N(0, 1) random numbers gives an Waveform where the resulting Waveform is now N(0, 0.1), as expected.
Bob Schor
Thanks a lot Bob! I'll try to implement your suggestion, haven't got it to run yet.
To clarify; I read a DAQ using DAQmx with the "Analog 1D Wfm, NChan NSamp" setting, where I have 13 channels at the moment and I pick 100 samples at a time. So what I want to do then is to compress each of the signals/ channels 100 times from 1000Hz to 10Hz. (I could of course pick 1000 samples as well but would prefer to run my "consumer loop" at 10Hz to have a decent rate of updates on the data, charts etc.
The problem I discovered now is that I was not extracting my 1 signal x 100 samples from my "main array". I don't even know what the array that DAQmx spits out actually looks like so I have bit of a hard time coming up with how to split it. The array size function tells me it's "13" so it seems my size is my channels. Where are the 100 samples? Hidden in each array element like a struct?
@KRICLA wrote:
@Bob_Schor wrote:
You start with a Waveform, and want to replace it with a Waveform that results from averaging over 100 points, resulting in a Waveform 1/100th the size (or at least that's what I assume you are doing). To have this work seamlessly, you'd want the size of the Waveform Y data chunk to be a multiple of 100, such as 1000 points. Then you'd simply peal off 100 points at a time, average them, and build a new waveform. There is no need for point-by-point averaging, since you (a) need to get the Waveform before you can process it, (b) this requires that all its points be available, so (c) there's no problem in "resizing" the array.
The processing is simple. Assume your waveform has size 100*m. Reshape the 1D Array into a 2D m-by-100 2D Array, pass this through a For Loop with a "Mean" function inside to convert it to an Array of size m consisting of the average of each sub-array of size 100. Use this to build a new Waveform, also (as Yamaeda noted) multiplying dt by 100. This example, based on a 1000 samples of a generated Waveform of N(0, 1) random numbers gives an Waveform where the resulting Waveform is now N(0, 0.1), as expected.
Bob Schor
Thanks a lot Bob! I'll try to implement your suggestion, haven't got it to run yet.
To clarify; I read a DAQ using DAQmx with the "Analog 1D Wfm, NChan NSamp" setting, where I have 13 channels at the moment and I pick 100 samples at a time. So what I want to do then is to compress each of the signals/ channels 100 times from 1000Hz to 10Hz. (I could of course pick 1000 samples as well but would prefer to run my "consumer loop" at 10Hz to have a decent rate of updates on the data, charts etc.
The problem I discovered now is that I was not extracting my 1 signal x 100 samples from my "main array". I don't even know what the array that DAQmx spits out actually looks like so I have bit of a hard time coming up with how to split it. The array size function tells me it's "13" so it seems my size is my channels. Where are the 100 samples? Hidden in each array element like a struct?
Yes, a Waveform is a Cluster (struct) with a t0, dT and an array of samples, that in your case should be 100 elements long. In Bobs picture above you can see the "Get waveform components" (or what it's called) as the 2nd block where you can access the different parts.
/Y
