RE: FFT and waveform

["Gofron, Kazimierz" <kgofron@bnl.gov>]

Ji Jan,

 

The FFTs that are currently computed in BEST system are quite noisy, even with NSLS2 beam current of about 200mA. Thus we are not finding these frequency computations as useful as we wished. Specifically difference between feedback on and off is not clearly visible with currently computed FFT inside BEST. Computing FFT outside of FPGA is sufficient, and more flexible.

 

Regards,

 

Kaz

 

From: Jan Marjanovic [mailto:j.marjanovic@caenels.com]
Sent: Tuesday, March 29, 2016 8:09 AM
To: Mark Rivers; Gofron, Kazimierz; 'tech-talk@aps.anl.gov'
Subject: Re: FFT and waveform

 

Hi Kaz,

there is also function to get FFT data already made available in low-level BEST library (see https://github.com/CAENels/best-epics-ioc/blob/master/bestApp/src/best_c_interface.h#L56). The easiest way to get the FFT data would be to add PV which calls this function. The getFFT() function currently relies on libfftw3, but we could also implement the FFT in FPGA if someday there is a need for that.


Best Regards,
Jan

On 3/28/2016 7:37 PM, Mark Rivers wrote:

Hi Kaz,

 

You should be able to just use the code in ADCore/ADApp/pluginSrc/fft.c.  It is only 132 lines of code, and has no external dependencies, so it can even run on VxWorks and RTEMS.  It contains these 2 functions:

 

void fft_1D(double data[], unsigned long nn, int isign);

void fft_ND(double data[], unsigned long nn[], int ndim, int isign);

 

These functions do the FFT in-place.  data[] is really a complex array, with each element of data[] consisting of 2 doubles, the first containing the real part and the second containing the imaginary part.

 

For your application the data from the electrometer will be put in the real part of data[],  the imaginary part will be set to 0.  isign should be 1.  nn is the array dimensions which must be a power of 2.  On return data[] will contain the complex FFT, with both real and imaginary values.

 

Mark

 

 

 

From: Gofron, Kazimierz [mailto:kgofron@bnl.gov]
Sent: Monday, March 28, 2016 11:01 AM
To: Mark Rivers; 'tech-talk@aps.anl.gov'
Cc: Jan Marjanovic (j.marjanovic@caenels.com)
Subject: RE: FFT and waveform

 

Hello Mark,

 

Indeed my inquiry is related to TetrAMM. The TetrAMM has two interfaces: Cat6 ethernet beamline network, and LC fiber (connected to BEST FPGA feedback system). The digital current measurement data flow through both interfaces simultaneously.

The 1G Ethernet communication is process by quadEM module and provides many areaDetector features including the upcoming FFT update.

 

In addition, CAENels released open version of the BEST ioc this month, and I have started using it.

GitHub: https://github.com/CAENels/best-epics-ioc

The view through BEST FPGA system allows information about implemented (piezo) feedback, in addition to passive monitoring that quadEM allows. However, the BEST IOC is missing some useful features like FFT. Adding FFT would enhance it.

 

Kaz

 

 

From: Mark Rivers [mailto:rivers@cars.uchicago.edu]
Sent: Sunday, March 27, 2016 1:27 PM
To: Gofron, Kazimierz; 'tech-talk@aps.anl.gov'
Subject: RE: FFT and waveform

 

Hi Kaz,

 

This is from the areaDetector/ADCore release notes for the upcoming R2-5 release.  These items are available now in the master branch on github.

 

****************************************

### NDPluginFFT

* New plugin to compute 1-D or 2-D Fast Fourier transforms.  It exports 1-D or 2-D

  NDArrays containing the absolute value of the FFT.  It creates 1-D waveform

  records of the input, and the real, imaginary, and absolute values of the first row of the FFT.

  It also creates 1-D waveform records of the time and frequency axes, which are useful if the 1-D

  input represents a time-series. The waveform records are convenient for plotting in OPI screens.

* The FFT algorithm requires that the input array dimensions be a power of 2, but the plugin

  will pad the array to the next larger power of 2 if the input array does not meet this

  requirement. 

*  The simDetector test application in

  [areaDetector/ADExample](https://github.com/areaDetector/ADExample)

  has a new simulation mode that generates images based on the sums and/or products of 4 sine waves.

  This application is useful for testing and demonstrating the NDPluginFFT plugin.

****************************************

 

 

This is from the areaDetector/ADExample notes for the upcoming R2-2 release.  These items are available now in the master branch on github.

 

****************************************

* Created a new driver in ADExample/exampleApp/ADCSimDetectorSrc.

  This driver generates times-series data for 8 signals as a 2-D array [8, NumTimePoints].

  The signals are common waveforms (sine, cosine, square wave, sawtooth, random noise, etc.)

 

  The driver was written in part to test the new NDPluginTimeSeries and NDPluginFFT plugins

  in ADCore R2-5. But it can also serve as a starting point for writing drivers that are used

  to collect time-series data like ADCs, waveform digitizers, etc. 

 

  There is a new iocs/iocADCSimDetector directory that creates an application for testing

  the driver.

* Added a new sine-wave simulation mode to simDetector.  This simulation mode constructs images

  as the sum or products of two sine waves in the X and Y directions.  It was designed in part

  to test the new NDPluginFFT plugin in ADCore R2-5.

****************************************

 

Mark

 

 

From: tech-talk-bounces@aps.anl.gov [mailto:tech-talk-bounces@aps.anl.gov] On Behalf Of Gofron, Kazimierz
Sent: Sunday, March 27, 2016 11:24 AM
To: 'tech-talk@aps.anl.gov'
Subject: FFT and waveform

 

Hi all,

 

My waveform (1024 current reading in [A], @1s) needs to be processed for frequencies using FFT method. The current values are in range between pA and mA ranges.

 

There is an existing tech-talk exchange on FFT from about 5 years ago that discuss FFT3, fftw, and waveProc.  

 

FFT3 based:

http://www.aps.anl.gov/epics/tech-talk/2011/msg01143.php

http://www.fftw.org/fftw3_doc/Complex-DFTs.html#Complex-DFTs

The github fftw development:

https://github.com/FFTW/fftw3

 

The waveProc is bit dated (2004?)

http://www.aps.anl.gov/epics/download/modules/index.php

 

Any newer solutions and suggestions?

 

Kazimierz Gofron

NSLS-II Controls Group

Brookhaven National Laboratory

(p) +1 (631) 344 5283

(f) +1 (631) 344 8085

(e) kgofron@bnl.gov

(w) www.bnl.gov/nsls2