Reports until 10:42, Tuesday 07 June 2016
H1 ISC (ISC)
carl.blair@LIGO.ORG - posted 10:42, Tuesday 07 June 2016 - last comment - 10:19, Wednesday 08 June 2016(27609)
ETMX PI transfer functions
[Rich,[Rich, Evan, Terra, Carl]
[Rich, Evan, Terra, Carl]
We measured the transfer function from the ETMX PI DAC LL channel to the LVLN outputs of D1500129.
The injection was made at the damping filter H1:SUS-ETMX_PI_OMC_DAMP_MODE4_DAMP_EXC.  With H1:SUS-ETMX_PI_OMC_DAMP_OUT_MTRX_1_4 = 1.  H1:SUS-ETMX_PI_ESD_DRIVER_PI_DAMP_SWITCH = ON.  H1:SUS-ETMX_BIO_L3_UL_STATEREQ = 1. H1:SUS-ETMX_BIO_L3_PI_ULLL_SW = ON. H1:SUS-ETMX_BIO_L3_LL_VOLTAGE_SW = OFF 
 
We used imaging to drive from 500Hz to 95kHz.  For example driving at 1kHz we read out a 1kHz, 64.5kHz and 66.5kHz signals.  In the first figure the transfer function is compared to a simulated transfer function derived from components:
>Gain of 80,000/2^17 as we used 80k counts
>sin(pi*f/fs)/(pi*f/fs) for the response of the DAC
>PI Bandpass filter transfer function from LIGO-S1500135-v1
>LVLN ESD amplifier transfer function from LIGO-D1500129-v1 (rough estimate from TF plot -> main error term)
The simulation is in agreement with the measurement other than at low frequency,  Not sure what happened there.  We will make more measurements with an uncertainty estimate from the measured SNR.
 
Friday we found a 20mv peak 1.26MHz noise (a clean sinusoid) on the ETMX LL channel, was not present on other channels.  Going back to investigate with Rich made this noise go away.
When the "ESD drive from DAC" input (J1B) is connected without the lowpass filter selected (PZ3 and PZ4) in the binary I/O there is ~2mV peak noise around 16kHz see second figure.
When the DAC input is connected without the lowpass filter selected in the binary I/O there is ~2mV peak noise around 16kHz. [Rich, Evan, Terra, Carl]
We measured the transfer function from the ETMX PI DAC channels to the LVLN outputs of D1500129
 
We used imaging to drive from 500Hz to 95kHz.  In the figure transfer function is compared to a simulated transfer function that has components:
Gain of 80,000/2^17 as we used 80k counts
sin(pi*f/fs)/(pi*f/fs) for the response of the DAC
PI Bandpass filter transfer function from LIGO-S1500135-v1
LVLN ESD amplifier transfer function from LIGO-D1500129-v1 (rough estimate from TF plot -> main error term)
 
The simulation is in reasonable agreement with the measurement.
 
We found at one point a 20uv peak 1.6MHz noise (a clean sinusoid) on the ETMX LL channel.  Going back to investigate with Rich made this noise go away.
When the DAC input is connected without the lowpass filter selected in the binary I/O there is ~2uV peak noise around 16kHz.Evan, Terra, Carl]
We measured the transfer function from the ETMX PI DAC channels to the LVLN outputs of D1500129
 
We used imaging to drive from 500Hz to 95kHz.  In the figure transfer function is compared to a simulated transfer function that has components:
Gain of 80,000/2^17 as we used 80k counts
sin(pi*f/fs)/(pi*f/fs) for the response of the DAC
PI Bandpass filter transfer function from LIGO-S1500135-v1
LVLN ESD amplifier transfer function from LIGO-D1500129-v1 (rough estimate from TF plot -> main error term)
 
The simulation is in reasonable agreement with the measurement.
 
We found at one point a 20uv peak 1.6MHz noise (a clean sinusoid) on the ETMX LL channel.  Going back to investigate with Rich made this noise go away.
When the DAC input is connected without the lowpass filter selected in the binary I/O there is ~2uV peak noise around 16kHz.
Images attached to this report
Comments related to this report
terra.hardwick@LIGO.ORG - 15:46, Tuesday 07 June 2016 (27630)
Link

We saw the noise at XEnd ESD LL again at ~1.28 MHz, 15mV pk2pk. Rich A. came out to have a look; believe the noise be real (a real oscillation in the hardware). 

On LR drive, we also saw the spectrum glitching (bouncing up and down) when driving at 15.5kHz; goes away with no drive signal (note we did not try other drive frequencies yet). Some beat making a low frequency component?

carl.blair@LIGO.ORG - 10:19, Wednesday 08 June 2016 (27649)
Link

The source of the 1.2Mhz oscillation was identified by opening the spare chasis and looking for marginally stable opamp stages.  
The stage that is marginally stable is U6 page 7 of D1500016.  It can be made to oscillate at ~600khz or ~1.2Mhz.  The stage is a configurable with the pole/zero bypass bit.  

When for example the H1:SUS-ETMX_BIO_L3_UL_STATEREQ control word is set to 2 the stage has a pole at 2.2Hz.  This is the normal low noise configuration.  In this configuration there is no 1.2MHz oscillation.
When this control word is set to 1 the stage is nominally a unitly gain stage.  In this configuration some channels (like UL UR LL and LR) have a Q of >5 at 1.2MHz and can be induced to freely oscillate.  This oscillation may be damped with a 30pF capacitor across R21.

As this oscillation is not a problem in the low noise configuration no changes will be made.  Testing PI channels should be performed with the H1:SUS-ETMX_BIO_L3_UL_STATEREQ control word set to 2.