Call ChnPeakDetect, finds "wrong" local peaks

[budist]

Hi,

I am using the following function on a measured "step"-function (values rise to a peak, then fall exponentially down to a certain level, then again rise to a peak and to a higher level and so on...):

Call ChnPeakDetect(Data.Root.ChannelGroups(1).Channels("Noname"),Data.Root.ChannelGroups(1).Channels("Noname_1"),"ChnY/Pos","ChnX/Ampl", "Peaks", 0, 3500)

I got 60000 values in the channel, with the peaks after each 5000 to 10000 values. 

But even with a threshold, Diadem gives me the wrong peaks, the number is correct, but the peaks are to low. Does the function give me some average-values? When I decrease the number of values (from 3500 to 3000) I get a little bit higher values.

What do I have to change or what did I wrong? Is it possible to get the max. local peaks in this way? I also tried "ChnPeakFind" but it didn't work out for me so far.

Thank you for you help!

[Brad_Turpin]

Hi budist,

Peak finding is something that humans do better than computers, especially when you're talking about teaching the computer how to find the peaks you're LOOKING at, based on your definition of what constitutes a peak.

The easiest way to answer your question would be to see your data set, and ideally to be able to play with it a bit.  Can you post it here or email it to me at brad.turpin@ni.com?

Brad Turpin

DIAdem Product Support Engineer

National Instruments

[budist]

Hi,

I've sent you the data. Peaks (column 2) in my case are the local maximum values. The peaks are "easy to look at" 

Best Regards

[Brad_Turpin]

Hi budist,

Here's the code I came up with to find your "step peaks".  Feel free to send me other "similar" data sets that the script doesn't work right on, and perhaps we can make it more robust, but it works great with the one data set you sent.

OPTION EXPLICIT
Dim n, P1, P2, Pn, Pmax, Group, Channels, XpeakChannel, YpeakChannel, RiseFormula, FallFormula
Const LowPassFiltFreq = 0.05
Const PeakAmplDivider = 5
Set Group = Data.Root.ActiveChannelGroup
Set Channels = Group.Channels
Call LowPassFilter(Channels(1), Channels(2), "Filtered", LowPassFiltFreq)
Call ChnDeltaCalc(Channels(3), "Delta")
Call ChnLinScale(Channels(4), Channels(4), CMax(Channels(2))/CMax(Channels(4)), 0)
Set XpeakChannel = Group.Channels.Add("Peak X", DataTypeFloat64)
Set YpeakChannel = Group.Channels.Add("Peak Y", DataTypeFloat64)
Pmax = Channels(4).Size
R1 = CMax(Channels(2))/PeakAmplDivider
RiseFormula = "Ch(Delta) > R1"
RiseFormula = Replace(RiseFormula, "Delta", """" & Channels(4).GetReference(eRefTypeIndexName) & """")
FallFormula = "Ch(Delta) < 0"
FallFormula = Replace(FallFormula, "Delta", """" & Channels(4).GetReference(eRefTypeIndexName) & """")
Call WndShow("VIEW")
Do
  P1 = ChnFind(RiseFormula, P2+1) ' rising edge
  IF P1 = 0 THEN
    P2 = Pmax
  ELSE
    P2 = ChnFind(FallFormula, P1+1) ' falling edge
    IF P2 = 0 THEN P2 = Pmax
  END IF
  IF P1 > 0 THEN
    n = n + 1
    Pn = 1 + FindLocalMaxRow(Channels(2), P1, P2)
    XpeakChannel.Values(n) = Channels(1).Values(Pn)
    YpeakChannel.Values(n) = Channels(2).Values(Pn)
  END IF
Loop Until P2 = Pmax

      
Sub LowPassFilter(ChnX, ChnY, ChnFilt, FiltFreq)
  IF FiltFreq > 0 THEN
    FiltLimit = FiltFreq
    FiltWndFct = "Rectangle"
    FiltType = "Low pass"
    FiltStruc = "IIR"
    FiltStyle = "Bessel"
    FiltZeroPhase = 1
    FiltCorrection = 1
    FiltDegree = 1
    Call ChnFiltCalc(ChnX, ChnY, ChnFilt)
  END IF
End Sub ' LowPassFilter()


Function FindLocalMaxRow(Channel, P1, P2)
  Dim k, Pn, PeakFormula
  FOR k = 1 TO 22
    StatSel(k) = "No"
  NEXT ' i
  StatSel(5)    = "Yes" ' Maximum value
  StatClipCopy  = 0
  StatClipValue = 0
  StatResChn    = 0
  Call StatBlockCalc("Channel", P1 & "-" & P2, Channel)
  PeakFormula = "ValEqual(Ch(Peak), StatMax)"
  PeakFormula = Replace(PeakFormula, "Peak", """" & Channel.GetReference(eRefTypeIndexName) & """")
  Pn = ChnFind(PeakFormula, P1)
FindLocalMaxRow = Pn
End Function ' FindLocalMaxRow()

Brad Turpin

DIAdem Product Support Engineer

National Instruments

[budist]

Hey Brad,

Thanks a lot for the code! It worked well on the example! I wanted to try it on multiple other data sets, but before that I got another problem and yet no solution:

I added a loop, but the subroutine wont save channel 4 in group 2 at  the second iteration.

Do you know why? I cant put the subroutine definition into the loop. But thats how the code is defined, I guess.

Can I work around this problem? The reason I create 3 groups, is that I get data files consisting of three data sets, which I gave you. So I want to split them up and get the results for each single data set in one group. Seems to be the easiest way in comparison to copying the channels and handling the channel numbers with an index (if that is possible...)).

Thanks again!

.....
Dim loop
for loop = 1 to Groupcount

'activates groups one after another ("if then", cause I couln't use index "loop" in Data.Root.ChannelGroups(loop).Activate())
if schleif=1 then
  Call Data.Root.ChannelGroups(1).Activate() 
  else 
    if schleif = 2 then
        Call Data.Root.ChannelGroups(2).Activate()        'activate group with index 2 in second loop iteration
    else
      if schleif = 3 then
        Call Data.Root.ChannelGroups(3).Activate() 
      end if
    end if    
end if

Set Group = Data.Root.ActiveChannelGroup        'select group "loop" at each iteration "loop"
Set Channels = Group.Channels             
....
...
your code up to beginning of subroutine definition
...
...
 END IF
Loop Until P2 = Pmax

next                                  'Loop end
      
Sub LowPassFilter(ChnX, ChnY, ChnFilt, FiltFreq)

  IF FiltFreq > 0 THEN
    FiltLimit = FiltFreq
    FiltWndFct = "Rectangle"
    FiltType = "Low pass"
    FiltStruc = "IIR"
    FiltStyle = "Bessel"
    FiltZeroPhase = 1
    FiltCorrection = 1
    FiltDegree = 1
    Call ChnFiltCalc(ChnX, ChnY, ChnFilt)              'doesn't save the channel ChnFilt in activated group, seems only in the group with index 1 , why? Didn't found an option for ChnFiltCalc to chose a group for saving the results in
       
  END IF
End Sub ' LowPassFilter()


Function FindLocalMaxRow(Channel, P1, P2)
  Dim k, Pn, PeakFormula
  FOR k = 1 TO 22
    StatSel(k) = "No"
  NEXT ' i
  StatSel(5)    = "Yes" ' Maximum value
  StatClipCopy  = 0
  StatClipValue = 0
  StatResChn    = 0
  Call StatBlockCalc("Channel", P1 & "-" & P2, Channel)
  PeakFormula = "ValEqual(Ch(Peak), StatMax)"
  PeakFormula = Replace(PeakFormula, "Peak", """" & Channel.GetReference(eRefTypeIndexName) & """")
  Pn = ChnFind(PeakFormula, P1)
FindLocalMaxRow = Pn
End Function ' FindLocalMaxRow() 

[Brad_Turpin]

Hi budist,

I'm not sure why you were getting an error, but I refactored my code a bit to support operating on all Groups in the Data Portal separately-- give this a try:

OPTION EXPLICIT
Dim Group
Const LowPassFiltFreq = 0.05
Const PeakAmplDivider = 5

FOR Each Group In Data.Root.ChannelGroups
  Call GroupStepPeaks(Group, LowPassFiltFreq, PeakAmplDivider)
NEXT ' Group


Sub GroupStepPeaks(Group, LowPassFiltFreq, PeakAmplDivider)
  Dim n, P1, P2, Pn, Pmax, Channels, XpeakChannel, YpeakChannel, RiseFormula, FallFormula
  Call Group.Activate
  Set Channels = Group.Channels
  Call LowPassFilter(Channels(1), Channels(2), "/Filtered", LowPassFiltFreq)
  Call ChnDeltaCalc(Channels(3), "/Delta")
  Call ChnLinScale(Channels(4), Channels(4), CMax(Channels(2))/CMax(Channels(4)), 0)
  Set XpeakChannel = Group.Channels.Add("Peak X", DataTypeFloat64)
  Set YpeakChannel = Group.Channels.Add("Peak Y", DataTypeFloat64)
  Pmax = Channels(4).Size
  R1 = CMax(Channels(2))/PeakAmplDivider
  RiseFormula = "Ch(Delta) > R1"
  RiseFormula = Replace(RiseFormula, "Delta", """" & Channels(4).GetReference(eRefTypeIndexName) & """")
  FallFormula = "Ch(Delta) < 0"
  FallFormula = Replace(FallFormula, "Delta", """" & Channels(4).GetReference(eRefTypeIndexName) & """")
  Call WndShow("VIEW")
  Do
    P1 = ChnFind(RiseFormula, P2+1) ' rising edge
    IF P1 = 0 THEN
      P2 = Pmax
    ELSE
      P2 = ChnFind(FallFormula, P1+1) ' falling edge
      IF P2 = 0 THEN P2 = Pmax
    END IF
    IF P1 > 0 THEN
      n = n + 1
      Pn = 1 + FindLocalMaxRow(Channels(2), P1, P2)
      XpeakChannel.Values(n) = Channels(1).Values(Pn)
      YpeakChannel.Values(n) = Channels(2).Values(Pn)
    END IF
  Loop Until P2 = Pmax
End Sub ' GroupStepPeaks()

      
Sub LowPassFilter(ChnX, ChnY, ChnFilt, FiltFreq)
  IF FiltFreq > 0 THEN
    FiltLimit = FiltFreq
    FiltWndFct = "Rectangle"
    FiltType = "Low pass"
    FiltStruc = "IIR"
    FiltStyle = "Bessel"
    FiltZeroPhase = 1
    FiltCorrection = 1
    FiltDegree = 1
    Call ChnFiltCalc(ChnX, ChnY, ChnFilt)
  END IF
End Sub ' LowPassFilter()


Function FindLocalMaxRow(Channel, P1, P2)
  Dim k, Pn, PeakFormula
  FOR k = 1 TO 22
    StatSel(k) = "No"
  NEXT ' i
  StatSel(5)    = "Yes" ' Maximum value
  StatClipCopy  = 0
  StatClipValue = 0
  StatResChn    = 0
  Call StatBlockCalc("Channel", P1 & "-" & P2, Channel)
  PeakFormula = "ValEqual(Ch(Peak), StatMax)"
  PeakFormula = Replace(PeakFormula, "Peak", """" & Channel.GetReference(eRefTypeIndexName) & """")
  Pn = ChnFind(PeakFormula, P1)
FindLocalMaxRow = Pn
End Function ' FindLocalMaxRow()

Brad Turpin

DIAdem Product Support Engineer

National Instruments

[jcamilleri]

Hi Brad,

Sorry for bumping an old thread but I have a 'similar' problem. I already posted my problem on the DIAdem forum, but today I found this thread and thought you might be able to help. I have a time trace of pressure from a sensor installed on a high-speed craft that was driven at high speeds in rough seas. The time trace consists of a large number of single impact events. I am using the Find Peaks tool to find the first 100 highest peaks. However the alogirthim is also picking up secondary peaks within the same impact event. Can you help me?

Regards,

Josef

[Brad_Turpin]

Hi Josef,

It looks from your screenshot like the secondary peaks always occur AFTER the primary peak chronologically-- is that always true?  Would you be able to assign a certain "dead window" time after each primary peak, so that any subsequent peaks in that time window could be categorically ignored?  Or might there be another, different primary peak that occurs before all the secondary peaks from the last primary are done?

Brad Turpin

DIAdem Product Support Engineer

National Instruments

[jcamilleri]

Hi Brad,

Thank you for your reply. I managed to sort out this problem with the help of Andreas (also a product support engineer) by first identifying all the 'impact events' and then searching for the peak within those events.

Regards,

Josef