NVGate Octave Analyzer

From OROS Wiki
Jump to navigation Jump to search

NVGate Octave plug in analyzer is especially designed for vibration and acoustic signal analysis needing 1/nth octave analysis and provides 1/1, 1/3rd, 1/12th and 1/24th digital filters that conform to the the latest acoustic standards.

It provides real time analysis and post analysis results

Moreover, the user can select input time domain filters with A, B or C law and time domain digital integrators and can apply on spectral results weights with A, B or C law.

A lot of averaging modes are available and to comply with acoustic standards, detectors also provide Fast, Slow, Impulse and LEQ averaging.

Octave Tracking is available with time, RPM, or any ΔV triggering.


Structure

Structure and Operation overview

The figure below shows the block diagram of the analyzer (for one channel) with octave set from 20 kHZ to 2,5kHz.


After amplification and antialias filtering, analog input signals are sampled at 51.2 kHz and converted by an analog to digital converter.

The sample frequency depends on the value of the High Filter Setting


Next, the input signals can be time weighted filters.


  • A, B and C weight filters : These filters available in acoustic frequency range (i.e. from 20 Hz to 20 kHz) satisfy requirements from last standards IEC 651 type 0 and IEC 804 type 0.
  • Any other NVGate filter need to be apply on Input Front end (real time) or Input Player (post analyisis).



After time domain filtering, the signals go to the digital 1/nth octave filter bank.

This filter bank is based on the 6th order digital band pass. A downsampling for low central frequency filters reduces computation load. The downsampling filters have a rejection greater than 90 dB at half of each new sampling rate and a ripple less than 0.007 dB in the useful bandwidth.

The filter range uses base 10 so that we get exact frequencies at 0.1 Hz, 1.0 Hz, 10 Hz, 100 Hz, 1 kHz and 10 kHz.


The following calculations are used to compute the central frequencies:



Number of filter and frequency range for each filter bank:

1/1 octave filter bank:

  • can have up to 11 filters (with 1 to 1000 frequency ratio),
  • covers range from 125 mHz to 16 kHz.

1/3rd octave filter bank:

  • can have up to 31 filters (with 1 to 1000 frequency ratio),
  • covers range from 100 mHz to 20 kHz.

For 1/12th octave:

  • the filter bank gets 4 filters for each useful bandwidth of 1/3rd filters, i.e. 124 filters,
  • covers range from 92 mHz to 21.8 kHz.

For 1/24th octave:

  • the filter bank gets 8 filters for each useful bandwidth of 1/3rd filter, i.e. 248 filters,
  • covers range from 90 mHz to 22.1 kHz.


The next step is the detector process for each 1/nth filter:

The filtered signal is input to a squaring module in order to get true RMS detector.

This module computes 1/N * Σx²n.

The instant averaging is always running and is based on the exponential averaging with a time constant equal to 1/fc where fc is the center frequency of corresponding filter. So each detector has its own time constant and the output fluctuation in the worst case is limited to +/- 0.4 dB with a sine input signal.

The averaging process provides a lot of modes (linear, exponential and dedicated to acoustical measurements) which are described in the 8.2.14.5 Settings | Analysis - Average Mode chapter.

A hold box allows to get Maximum and Minimum spectra during one measure.

A stabilization delay is implemented in order to ignore and suppress the transient response of passband filters. It is automatically taken into account after any change of input setup. It is equal to 5 periods of the lower frequency filter for 1/3rd octave and octave filters. This delay is four times greater for 1/12th octave filters (i.e. 20 periods of the lower frequency filter) and eight times greater for 1/24th octave filters (i.e. 40 periods of the lower frequency filter).For example, if lower frequency filter is centered at 1 Hz, then the stabilization delay is equal to 5 seconds for octave and 1/3rd octave, 20 seconds for 1/12th octave and 40 seconds for 1/24th octave. During this delay the detectors are inactive.

The filter bank continuously runs and instant averaging is always available (excluding time period for filter stabilization).

In linear modes, trigger events are used to start and/or stop energy computation in detectors and store associated results in the spectrum waterfall memory.

In exponential modes, detectors are always running and trigger events are only used to store associated results in the spectrum waterfall memory.

NVGate 1/n octave plug in

This plug-in computes 1/1, 1/3, 1/12 and 1/24 octave analysis based on order 3 digital filters and RMS detectors.

The filter bank complies with:

IEC 1260 class 1 andANSI S1.11-1986

Mid band frequencies are based on powers of ten.

Available results:

FilterNumber = ((1/Mode) * (ln (High/Low) / ln 2)) + 1

Where:

"Mode" is the value of the 1/n octave/CPB filters/Mode setting: 1 for octave, 1/3 for 1/3 octave…

"High" and "Low" are respectively the values of the 1/n octave/CPB filters/Upper central freq and 1/n octave/CPB filters/Lower central freq settings

For example if mode = 1/3 octave, High =7.94 kHz and Low = 7.94 Hz then FilterNumber = 3 * (ln 1000 / ln 2) +1 = 30.897 => 31 filters

Type Size Dimension Domain Save
1/n octave FilterNumber 2D spectral Yes
Avg. 1/n octave FilterNumber 2D spectral Yes
Min Avg. octave FilterNumber 2D spectral Yes
Max Avg. octave FilterNumber 2D spectral Yes
Overall level 1 pt 1D level Yes
Overall Weighted 1 pt 1D level Yes


  • 1/n octave: This result is always available. It displays the output of the RMS detectors that analyzes the output of each filter from the filter bank.
  • Avg. 1/n octave: This result is available when the 1/n Octave plug-in is in "Running" state. It displays the result of the averaging of the 1/n octave result.
  • Min Avg. octave: This result is available when the 1/n Octave plug-in is in "Running" state. It displays the minimum value of each band of the Avg 1/n octave result. Restarting the plug-in will reset this result.
  • Max Avg. octave: This result is available when the 1/n Octave plug-in is in "Running" state. It displays the maximum value of each band of the Avg 1/n octave result. Restarting the plug-in will reset this result.
  • Overall level: This result is available when the 1/n Octave plug-in is in "Running" state. It displays the overall level of the signal between the lower and upper frequency. It is computed before the filter bank, thus it is not the addition of the values of the 1/n octave result.
  • Overall Weighted: This result is available when the 1/n Octave plug-in is in "Running" state. It displays the overall level of the weighted signal between the lower and upper frequency. The weighted filter is chosen by the 1/n octave/CPB filters/Global Level Weighting setting. It is computed before the filter bank, thus it is not the addition of the values of the 1/n octave result.

Computation SPUs:

Bandwidth Fdec Reso SPU/Channel
for Real-time
25.6k 1 1/3rd 4
20k 1 1/3rd 3
12.8k 1 1/3rd 2
10k 1 1/3rd 1,5
20k 1 1/3rd 3,0
10k 2 1/3rd 2,0
5k 4 1/3rd 1,25
20k 1 1/1 1,5
20k 1 1/3rd 3
20k 1 1/12th 6
20k 1 1/24th 12


Sampling Frequency: set in Front-End/Inputs settings/Input sampling

1/N Oct Bandwidth: set in OCT/FFT analysis/range

Channel

Contains the settings related to the source input.

  • Source: input source to be analyzed. It may come from the Front-end input or from the Player in post analysis mode (see the post analysis chapter). In post analysis mode, tracks with a signal bandwidth lower than the Upper central frequency cannot be plugged.
Trigger

Contains the settings related to triggering events and how to start and stop signal computation.

  • Start: defines the event to start the analysis. The user can choose any event among the list of defined events. By default only the Free run and Manual events are available. The user can define another event in the "Event Definition" shared resources and then use this event for the "Start" condition.
  • Stop: defines the event to stop the analysis. The user can choose any event among the list of defined events. By default only the Free run and Manual events are available. The user can define another event in the "Event Definition" shared resources and then use this event for the "Stop" condition.
  • Repeat: This setting allows selecting a condition for the plug-in to be restarted. If an event is selected as a Start event, the option New start is available, and when this mode is selected each Start event restarts the plug-in. The mode End of averaging is available when the value of the 1/n Octave x/Average/Type setting is different from the Exponential mode, it restarts the plug-in when the averaging is finished. It is set to "Off" except for linear average when it is set to "End of averaging". In this case, the "Avg duration" gives the periodicity of the average restart.
Repeat Description
Off no repeat. The measurement is stopped at the end of averaging.
New start The measurement restarts when the start event occurs. It is enabled if Start is different from Free run.
End of averaging The measurement restarts at the end of averaging.


Hidden/fixed: the average type sets the authorized repeat modes.

Type Repeat mode
Linear Off / New start / End of Averaging
Repeat Set to End of averaging
Short LEQ (1/8s or 1s) Set to End of averaging

Other types
Set to Off


The following scheme is a description of the two different repeat modes:

- Mode Repeat on ’end of averaging’:

The ’Run’ corresponds to the start of the analyzer. The start is the beginning of average duration. At the end of the first average, the second average will start and so on till the end of the integration time (even if the last short time integration duration is not over). Each averaging has the same size, except the last that can be smaller.

- Mode Repeat on ’New start’:

The ’Run’ trigger the analyzer, but the average begins at the start event (set in the event definition). This averaging stops at the end of average duration or at the stop event. Then the next averaging is waiting for the new start event to occur.

In case a new event occurs during a previous averaging, this start event won’t be effective because the previous average duration is not over, you can only set other event after the end of the previous one.

Average

Contains the settings related to the type of averaging of the signals to be computed.

  • Type: the average type.
Type Description
Linear Arithmetic average of the instantaneous 1/n octave during the "Avg duration"
Repeat Performs linear averaging over "Short duration" time and finally computes a linear average of all "Short duration" results over "Avg. duration" time

Refresh is performed every "Short duration" with the display if the corresponding linearly averaged spectrum over "Short duration" time.
The final averaged spectrum corresponds to the linear averaged spectrum over "Avg. duration" (equivalent to the linear average of each "Short duration" result).

The event "Oct result available" (to be used for instance as a trigger in the waterfall) corresponds to the Avg. 1/n oct spectrum after each "Short duration" calculation.
Exponential Works with the same time constant on every detector. It is equivalent to a RC filter following the squaring module. Performs continuous exponential averaging of instantaneous 1/n octave spectrum over "Exponential time constant" value.

Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency)

The event "Oct result available" corresponds to the periodicity (1/(fs* 2,56))*256 with fs the input sampling frequency).
Constant BT 0.2 dB Performs exponential averaging with averaging time in each 1/n octave band inversely proportional to the frequency so that the product B*T is constant.
For "Constant BT 0.2 dB": B*T = 500
with B the center frequency of 1/n octave band and T the corresponding integration time.
This averaging setting gives a standard error <0.2dB

Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency)

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).
Constant BT 0.5 dB Performs exponential averaging with averaging time in each 1/n octave band inversely proportional to the frequency so that the product B*T is constant.
For "Constant BT 0.5 dB": B*T = 100
with B the center frequency of 1/n octave band and T the corresponding integration time.
This averaging setting gives a standard error <0.5dB

Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency)

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).
Constant BT 1 dB Performs exponential averaging with averaging time in each 1/n octave band inversely proportional to the frequency so that the product B*T is constant.
For "Constant BT 1 dB": B*T = 25
with B the center frequency of 1/n octave band and T the corresponding integration time.
This averaging setting gives a standard error <1dB
Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency)

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).
Constant BT 2 dB Performs exponential averaging with averaging time in each 1/n octave band inversely proportional to the frequency so that the product B*T is constant.
For "Constant BT 2 dB": B*T = 5
With B the center frequency of 1/n octave band and T the corresponding integration time.
This averaging setting gives a standard error <2dB

Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency)

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).
Short LEQ 1/8s Corresponds to a linear averaging with a fixed "Short duration" of 1/8s (equivalent to Repeat averaging mode with non selectable "Short duration" parameter).
New result available every 125 ms for display and for the Waterfall.

The event "Oct result available" corresponds to the Avg. 1/n oct spectrum after each "Short duration" calculation (every 125ms).

This average mode is not available for frequency ranges below 5kHz.
Short LEQ 1s Corresponds to a linear averaging with a fixed "Short duration" of 1s (equivalent to Repeat averaging mode with non selectable "Short duration" parameter).
New result available every second for display and for the Waterfall.
Note: The tenth second measure corresponds to the average of the entire measurement (not the average for the tenth second). Moreover if you stop the measurement before the end, the last measurement would not be displayed.

The event "Oct result available" corresponds to the Avg. 1/n oct spectrum after each "Short duration" calculation (every 1s).

This average mode is not available for frequency ranges below 5kHz.
Fast Corresponds to an exponential averaging with a fixed "Exponential time constant" of 0.125s.
Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency)

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).

This average mode is not available for frequency ranges below 5kHz.
Slow Corresponds to an exponential averaging with a fixed "Exponential time constant" of 1s.
Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency).

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).

This average mode is not available for frequency ranges below 5kHz.
Impulse Corresponds to an exponential integration with a raise Exponential time constant of 35 ms and a fall Exponential time constant of 1500ms.

Refresh is performed when new instantaneous spectrum is available (periodicity: (1/(fs*2,56))*256 with fs the input sampling frequency).

The event "Oct result available" corresponds to the periodicity ((1/(fs*2,56))*256 with fs the input sampling frequency).
This average mode is not available for frequency ranges below 5kHz.


Hidden/fixed: the last five types are enabled only if the Upper central frequency is upper than 6.3 kHz.

  • Avg. duration: defines the long term average duration time in second. See the table below for more information. If you change the short duration, the average duration will be automatically adjust to the closest value multiple of short duration.
  • Short duration: defines the short term average duration time in second. See the table below for more information.
  • Exponential time constant: defines the average duration time in second for the exponential averaging mode. See the table below for more information.

Hidden/fixed:

Type Avg. duration Short duration Exponential time constant
Linear Visible not used not used
Repeat visible visible not used
Exponential not used not used visible
Cst BT not used not used not used
Short LEQ1/8s visible set (1/8s) not used
Short LEQ 1s visible set (1s) not used
Fast not used not used set (1/8s)
Slow not used not used set (1s)
Impulse not used not used set (0.035s)
CPB filters

Defines the frequency range and the resolution.

  • Autobandwidth: manages the analysis and recording bandwidths automatically. The Inputs selection window allows the use of 2 different sampling rates for the dynamic inputs. It gathers inputs with the same physical quantity into groups, maintaining the same sampling into each group. When inputs are associated with the analysis plug-in, it adjusts its analysis bandwidth to match the inputs ones. Mixing input bandwidths in one analysis plug-in lead to set its bandwidth to the lowest one. Autobandwidth is set by default . When it is enabled in the plug-in, these settings "switch to informative status"
  • Lower central freq: the central frequency of the lower band of the frequency range. The bands considered are octave bands (for the octave mode) and 1/3 octave bands (for the other modes).

The user enters a value, which is adjusted to the closest central frequency. The bandwidth between the Lower and Upper central frequencies must be lower than or equal to eleven filters (considering octave bands) and 31 filters (considering 1/3 octave bands). So, if this bandwidth increases, then the Upper central frequency is automatically reduced.

  • Upper central freq: the central frequency of the upper band of the frequency range. The bands considered are octave bands (for the octave mode) and 1/3 octave bands (for the other modes).

The user enters a value, which is adjusted to the closest central frequency. The bandwidth between Lower and Upper central frequency must be lower than or equal to eleven filters (considering octave bands) and 31 filters (considering 1/3 octave bands). So, if this bandwidth increases, then the Lower central frequency is automatically increased.

  • Mode:
Mode Description
Octave Octave digital filter
1/3 octave 1/3 octave digital filter
1/12 octave 1/12 octave digital filter
1/24 octave 1/24 octave digital filter
  • Global Level weighting:

Weighted overall levels of the1/n Octave Plug-in are now computed on the time domain (weighting filter and detector). Processing weighting in the time domain provides accurate measurement for non-stationary signals (impulsive).

The type of time domain weighting (A or C) is selected in the ASB setting 1/n Oct / CPB filters / Global level weighting.

These overall values can be displayed in view meter.

/n Octave status

All statuses are available to add to the control panel

/N oct

The current plug-in status is synthesized in a special progress-bar. This progress bar is automatically displayed in the ’control panel’ when the plug-in is active (i.e. as soon as at least 1 input is connected to the Sync Order plug-in). This status is called ’1/N Oct’ and it is available in the status ASB tree (see customize control panel).

This status displays the type of averaging between brackets (i.e. Lin, Rep, Leq, CBT, Fast, Slow, Ipls or Exp), the plug-in state (Run, Paused, Stop) and the real-time status.

The way the status bar is displayed depend on the mode selected:

  • Linear, Repeat, Short Leq: display the Average Duration. The status bar is displayed from left to right. The text displayed indicates the state of the acquisition (Run, Paused or Stop) with the time.
  • Cst BT, Fast, Slow, Impulse: the status bar is a ’snake’ type; the text displayed indicates the state of the acquisition (Run, Paused or Stop).
  • Exponential: the status bar is displayed from right to left with flashing until the end of the acquisition. The text displayed indicates the exponential averaging time.

The color of the background bar and of the text depends on real-time status:

  • Green background and white or black text: acquisition in real-time.
  • Red background and white or black text: the current acquisition is not real-time (current block is not analyzed).
  • Green background and red text: the current acquisition is real-time, but since start not all the blocks were analyzed.
Trigger state

The following scheme describes the different states of the measurement:

Count

This status displays:

  • The average time in linear, short Leq and repeat modes.
  • The exponential time constant in exponential, Fast, Slow, Impulse modes.
  • The integration time T in constant band tracking modes.

Note that in linear mode, if repeat were on end of averaging, the count would restart at the end of averaging until stop event. If repeat were on new start, the count is set to zero at the end of averaging waiting for a new trigger.

In Short Leq mode, the count is stopped at the end of time duration (the repeat mode is made on short duration).

Analysis overload

This status displays if during the acquisition, there were an amplitude overload (or not) of the analyzed inputs.

Time to stabilization

Until this time to stabilization is over, results are not available. All filters have a stabilization time and at any change of settings, filters need this time before the results are available. The lower the frequency is the longer this stabilization is.

The filters stabilization time is given by:

Time (s) = x * 5 / lower central frequency

Where x depends on the mode:

Mode x
Octave 1
1/3 octave 1
1/12 octave 4
1/24 octave 8

If there were no change between two run, no stabilization would be necessary. But if a new plug-in is connected, the system is reinitialized and then there is a time to stabilization.