At the end of this reported activity the plan was to restore the INJ_MAIN automation node (it was put in pause before) but Matthieu was reporting to me he had problem on his thinlinc session with the automation client not correctly displayed.

I was not able to reproduce the problem from remote (it may have been something specific on the thinclient box in Control Room) and I could easily put the node back in exec and drive it to IMC_RESTORED (see attached screenshot)

This morning we kept on working on Fmoderr.
We tuned the phase of the RF numerical demodulation in ACL (see figure 1)
The new phases are the following:
- 6 MHz : 2.59
- 8 MHz : 2.83
- 56 MHz : 1.14

Then we did a scan of the MC in order to do the calibration of the 56 Mhz signal that we would like to use as the new error signal.
The 1111 Hz line was set to 0.1 V which is the amplitude which is used in the automation. For the tests of Friday it was set to 0.2. 
The scan started at 12:35:10 UTC and last 300 s.
 

Figure 2 presents the evolution of those three digital error signals and the analogical one (currently used, 8 MHz) while scanning MC_F0_Z. In blue the raw signals, in orange the 1s trend computed with matlab.
 

Figure 3 is a zoom of those scans around the zero crossing and a linear fit. The parameters of the fits are given in the title of the plots.
Especially we can notice that the MC position for which the error signals are crossing zero are relatively close:
- 6 MHz : -61
- 8 MHz : -68
- 8 MHz analog : -67
- 56 MHz : -72
 

Visually we can notice that the 56 MHz is much cleaner than the other signals. We computed the FFT of the signals over the scan (figure 4). The useful information is in the 1e-2 1e-1 band while the higher frequencies is noise. There is more than 1 order of magnitude for the 56 MHz, a quite low SNR for the 8 MHz both analogically and numerically demodulated and a SNR close to 1 for the 6 MHz.

There are still some points to be understood.
In particular why the slope of the 56 is negative while the two other are positive? It might come from the 1111 demodulation in ACL.
Why the Fmoderr at 6 MHz is almost not given any information? It was much less noisier during the scan of friday (see figure 5, MC_F0_Z scan of different durantion and amplitude...). The only two parameters that have changed are the RF demodulation phase and the amplitude of the 1111 line.

Side note, the bad behavior of the 6MHz is not due to a demodulation phase mistuning, see figure 1.

The figure attached show a better plot for what concern the 6 MHz loss of information.
As metionned in the previous entry the only two parameters that have changed were the amplitude of the 1111 line and the demodulation phase.

We suspected that the phase tuning that we have done was spoiled by the RAM. The demodulated signal (DAQ_FMOD_6MHz) could indeed also contain a component due to RAM.
To check it, for each signal, we computed the TF between the I and Q signal (DAQ_FMOD_6MHz_I .... ) and the 1111 line. The Fmoderr is sensitive to the frequency perturbation while the RAM is not (less).

Here below are the results: 

So there is indeed very few information in the 6 MHz_I signal.
To be noted that there was quite a low coherence between the signal and the 1111 line. We should increase the line to get more coherence, redo the measurement and maximize the Fmoderr information in the I signal.
We should also investigate on the RAM.

About the problem on BsX_ML_PZT mentionned in the log entry 56520 :

###

BsX_ML_PZT

While switching on the 1111 Hz line we noticed that it was not visible in the correction send to the PZT of the ML. 
We had to increase it quite a lot to be able to see it.
We investigated a bit and realized that there is no coherence between the error signal and the correction. The monitoring channel reading the PZT has a problem that will have to be adressed. (figure 3)

###

This mornig we switched the two signals INJ_ML_PZT (same than BsX_ML_PZT) and INJ_IMC_TRA_6MHz to see if the problem was coming from the signal itself or from its acquisition.
The acquisition is ok (figure attached) so it seems that the problem is coming from the monitor signal coming from the rampeauto.

The digital double demodulation done for FModerr is using the signal DAQ_FMOD_Out0 which comes from the photodiode INJ_IMC_TRA.
By looking at the signal, it appears that we can already see some lines at 6 8 and 56 MHz while this a power signal directly coming from the photodiode. Those lines could be associated to amplitude modulation (RAM).
However by taking a look at a GPS time when there was no beam going towards the injection (Chrroc installation) the lines are still present (figure 1). So it seems that the signal gets polluated by the RF frequencies between the photodiode and the demodulation mezzanine.
Figure 2 3 4 are zooms on the 6 8 and 56 MHz lines. One can see a difference of amplitude with and without optical signal for the 6 MHz and the 8 MHz. This can be related to RAM.

This morning we did some small tests to investigate on the DAQ_FMOD_out0 pollution.
The signal sent towards the demodualtion mezzanine is a derivation from the signal which is sent towards the analog demodualtion at 8 MHz.
We unplugged the cable at the level of the demodualtion mezzanine and we observed a normal noise floor (figure 1). This confirms that the signal is polluted before.

We unpluged the INJ_IMC_TRA signal arriving at the level of the derivation before the analog demodulation and we could still observed the peak at 8 MHz. This confirms that the signal is polluted at 8 MHz by the analog demodulation.
We unplugged the 8 MHz LO signal used for the analog demodualtion and we could still observed the peak at 8 MHz. This confirms that the signal is anyway polluted by the 8 MHz on its way towards the demodualtion box.

We were suspecting that the INJ_IMC_TRA get polluted by the RF frequencies through the lemo-3 cable used to bring the signal from the photodiode towards the rack where the demodulations are done in piscina.
This morning we went in laser lab and substitute this lemo-3 cable by a BNC cable that was already in place and not used between EIB and the piscina (phase camera CH2 dist, see photo).

As a consequence we have more power in the INJ_IMC_TRA_6MHz signal. This might be due to a better phase tuning. This demodulated signal is now cleaner (see figure attached).
The signal INJ_IMC_TRA seems more polluted around the harmonics of the 50 Hz. It seems that it comes from the cable itself since the PSTAB_PD1_AC signal remained the same.

The change in the digital demodualtion signals still need to be analyzed.