Reports 1-1 of 1 Clear search Modify search
AdV-DAQ (Data Acquisition and Global Control)
casanueva, masserot, letendre, pacaud - 16:32 Tuesday 06 February 2018 (40545) Print this report
SSFS - DSP low noise filter implemetation

We deployed the so-called low noise filter implementation designed by D.Passuelo :

  • upgrade of  the DSP code
  • upgrade of the DaqBox server with v12r2
  • upgrade of Acl  v050p12

The DAC ouput is disconnected  with the master laser  .

A long test is in progress with :

  • a injected noise built wit the SSFS_noise_alt filter

ACL_FILTER_SET        SSFS_noise_alt    1    7.53743e-09    1    12
ACL_FILTER_ZEROS    SSFS_noise_alt    5000    0.5
ACL_FILTER_POLES    SSFS_noise_alt    300        0.5

  • the SSFS_B4_GAIN set  to 0
  • the SSFS_ON flag set to 1 by the ACL_ADC_CH

ACL_ADC_CH    SSFS_ON                "au"        1    SSFS_DSP_FREQ    LSC_SWITCH_SSFS_ON    1        1        "NONE"

  • the  DSP correction output built with the   "O3_SSFS_B4flt_base * O3_SSFS_B4flt_boost3"  filters provided by Julia :

ACL_FILTER_SET        "O3_SSFS_B4flt_base"    2    -7.8    10000    10
ACL_FILTER_ZEROS    "O3_SSFS_B4flt_base"    0        0.
ACL_FILTER_POLES    "O3_SSFS_B4flt_base"    14        0.
ACL_FILTER_POLES    "O3_SSFS_B4flt_base"    50000    1.50000
 

ACL_FILTER_SET        "O3_SSFS_B4flt_boost3"    1    1        10000    2
ACL_FILTER_ZEROS    "O3_SSFS_B4flt_boost3"    1000    1
ACL_FILTER_POLES    "O3_SSFS_B4flt_boost3"    50        0.70000
ACL_FILTER_ZEROS    "O3_SSFS_B4flt_boost3"    1500    1.
ACL_FILTER_POLES    "O3_SSFS_B4flt_boost3"    50        1.
ACL_FILTER_ZEROS    "O3_SSFS_B4flt_boost3"    1650    5.
ACL_FILTER_POLES    "O3_SSFS_B4flt_boost3"    1650    25.

 

The goal is to chech  the stabilty of the DSP filter output running with the low noise computation algorthim

Comments to this report:
masserot, rolland - 17:19 Thursday 22 February 2018 (40662) Print this report

Acl v050p12 /virgoStaging

The first plots show over 2 days the evolution on the digital noise when we use the low noise filter implementation with the boost filter inside the SSFS DSP:

  • the filter has the following bi-quad implementation in the S space:
    • S space: "ZERO_AdV"(0)-1Hz Q:0 - "POLE_AdV"@(51712.6)50000Hz Q:1.5  - Gain: 1 - Coef 1 0 0 / 1 2.05179e-06 9.47214e-12
    • S space: "ZERO_AdV"(0)0Hz Q:0 - "POLE_AdV"@(1650.06)1650Hz Q:25  - Gain: 1 - Coef 0 1 0 / 1 3.85816e-06 9.30339e-09
    • S space: "ZERO_AdV"(1000.01)1000Hz Q:1 - "POLE_AdV"@(50)50Hz Q:0.7  - Gain: 1 - Coef 1 0.000159153 2.53296e-08 / 1 0.00454728 1.01321e-05
    • S space: "ZERO_AdV"(1500.04)1500Hz Q:1 - "POLE_AdV"@(50)50Hz Q:1  - Gain: 1 - Coef 1 0.0001061 1.12572e-08 / 1 0.0031831 1.01321e-05
    • S space: "ZERO_AdV"(1650.06)1650Hz Q:5 - "POLE_AdV"@(14)14Hz Q:0  - Gain: 1 - Coef 1 1.92908e-05 9.30339e-09 / 1 0.0113682 0
  • one can see that  the usal jumps on the SSFS_B4_DigitalNoise_corr_{min,max}  channels have disappears
  • nertheless the difference between the SSFS_B4_DigitalNoise_corr_min and the SSFS_B4_DigitalNoise_corr_max increase in the 2 days of  tests
  • and the residual digital noise  SSFS_B4_digitalNoise_DS_{min,max} remains at the same level.

Looking at the fullband channel (500kHz), we found that

  • a line at 250kHz appears at the DSP filter output (low noise implementation) and at the RTPC filter output (standard filter implementation) too
  • the amplitude of this line increase slowly with the time
  • and that  the line frequency is always half of the input channel  frequency.

 

Acl trunk /olusers/masserot launched only on the SSFS_Ctrl server.

The second plot show  over 6days  the evolution on the digital noise, in the same conditions as previously, with the zero ordered in other way inside the bi quad:

  • the filter has now  the following bi-quad implementation in the S space:
    • S space: "ZERO_AdV"(0)0Hz Q:0 - "POLE_AdV"@(51712.6)50000Hz Q:1.5  - Gain: 1 - Coef 0 1 0 / 1 2.05179e-06 9.47214e-12
    • S space: "ZERO_AdV"(1000.01)1000Hz Q:1 - "POLE_AdV"@(1650.06)1650Hz Q:25  - Gain: 1 - Coef 1 0.000159153 2.53296e-08 / 1 3.85816e-06 9.30339e-09
    • S space: "ZERO_AdV"(1500.04)1500Hz Q:1 - "POLE_AdV"@(50)50Hz Q:1  - Gain: 1 - Coef 1 0.0001061 1.12572e-08 / 1 0.0031831 1.01321e-05
    • S space: "ZERO_AdV"(1650.06)1650Hz Q:5 - "POLE_AdV"@(50)50Hz Q:0.7  - Gain: 1 - Coef 1 1.92908e-05 9.30339e-09 / 1 0.00454728 1.01321e-05
    • S space: "ZERO_AdV"(0)-1Hz Q:0 - "POLE_AdV"@(14)14Hz Q:0  - Gain: 1 - Coef 1 0 0 / 1 0.0113682 0 - B4_Corr_RTPC_BK1:O3_SSFS_B4flt_base*O3_SSFS_B4flt_boost3 - "double_II"(0) -  resetReq 0
  • in this case the difference between the SSFS_B4_DigitalNoise_corr_min and the SSFS_B4_DigitalNoise_corr_max  remains constant during the 6 days of tests

 

The third plot show the FFT spectrum of

  • the input filter channel
  • the DSP 40Bit output filter 
  • and in the third graph the standard SSFS_dac ouput(black), the digital noise correction(orange) and in the remaing noise (green)

 

The new bi-quad implementation is:

  • less noisy
  • and stable over the time

This new bi-quad repartition will be available in the next Acl release .
 

Images attached to this comment
Search Help
×

Warning

×