Yes, things have changed.  Please wait until I have things cleaned up before working on models.  I'll explain what the new setup is.

 Here are the issues that I found not quite accurate in the old oaf code:

1. There is no need to calculate the norm of the witness signal every time from zero: 

norm += (*pBufAdapt) * (*pBufAdapt); // add to the norm 

Every step the witness signal vector is the same except the first and last values

wit[i].norm += histAdpt[nCoeff]*histAdpt[nCoeff] - histAdpt[0]*histAdpt[0];

 This step will reduce the number of multiplications and summations from 3*M/k to 2*M/k, M - filter length, k - downsample ratio.

2. Old code filter corrects filter coefficients with a delay equal to k=downsample ratio (pretty big):

witness       o o o o o o o o o o o o o o o o o o o o o

error           o o o o o o o o o o o o o o o o o o o o o

We want the filter to work at green points and skip red points computing output and correcting coefficients at this time (downsample ratio in this example is 4). Old code

• grabs error signal
• calculates output during next k-1 red points and 1 green point
• corrects coefficients using this error during next k-1 red points and 1 green point

But LMS algorithm should correct coefficients according to the latest error. As we calculate output and correct coefficients before the latest error signal will be available, we should change the order:

• grabs error signal
• corrects coefficients using this error during next k-1 red points and 1 green point
• calculates output during next k-1 red points and 1 green point

This scheme is completely equivalent to the ordinary LMS algorithm, as now we correct coefficients according to the latest error signal and so do not add any delay.

3. So long soft start is probably not needed for the LMS filter, it makes the filter to converge longer 

// modify adaptation gain for soft start

    if( state.iWait < state.nFIR )

    {

      adaptGain = 0.0;

      decayRate = 1.0;  // clear FIR coeffs after reset

    }

As far as I understand this is done to prevent the filter from huge coefficients variations in the beginning when norm is not calculated yet. Instead we can just introduce some small

epsilon = 10.0;

to prevent the filter from divergence in the beginning 

delta = mu * error / (wit[i].norm + epsilon);

Though some soft start might be needed by not so long - it will take several minutes before the adaptation gain will take it's specified value. 

Today I tried to make the lms filter to work online. I played around with the signals (GUR1_X and MC_F) to pre-whiten them and in the end the following configuration worked out:

1. mu = 0.03, tau = 1e-5, downsample=8, nCoeff = 4000, delay = 5 (sample-and-hold delay is not included in the new code, it should be added here!)

2. witness pass: AA32 = cheby1("LowPass", 4, 1, 32) AND 0.1:0

3. witness adaptation path: AA32 AND AI32 = cheby1("LowPass", 4, 1, 32) AND 0.1:0

4. error path: AA32 AND 0.1:0 AND anti_1Hz. Before I added anti_1Hz filter oaf did nothing. This filter tries to approximate the actuator transfer function. Note, it is not in the witness adaptation path. This is some sort of whitening.

5. correction path: AI32, gain = -1

Convergence time ~ 5 mins. The performance of the filter is far not perfect compared to the offline implementation. But it deals with a stack though.

I worked a bit on the PSL table today

It isn't clear to me in the drawing where the Agilent is during this measurement. Over 40m of cabling, the loss of signal can be a few dB, and considering we don't have a whole lot of signal in the first place, it may be better to send the stronger RF signal (i.e. Marconi pickoff) over the long cable rather than the weak beat signal from the Transmission photodiode. 

[Jenne, Zach, Kiwamu]

 We made some efforts to lock the X arm cavity with the infrared beam.

We eventually obtained the PDH signal from a photo diode at AS port, we are still in the mid way of the lock.

The PDH signal now is going into c1lsc's ADC.

We have to make sure which digital channel corresponds to our PDH signal,

(what we did)

- split the LO signal, which comes from a Marconi, just before the EOM into two path.

     One is going to the EOM and the other is going to the AP table for the demodulation, The driving frequency we are using is 11MHz.

- put RF amplifiers to make the RF signal bigger. The raw signal was small, it was about ~-50 dBm in the spectrum analyzer.

   So we connected two ZLN amplifiers. Now the RF signal is at about 0 dBm

- connected the LO and RF signals to a mixer.  Additionally we put a 9.5-11.5 MHz bandpass filter at the LO path since there was some amounts of 29.5MHz due to the RF reflection at the EOM resonant box.

     After a low pass filter by SR560 the signal shows typical PDH behaviors.

- strung a BNC cable which connects the demodulated signal and c1lsc.

    In order to connect the signal to the current working ADC, we disconnected AS166_I from a whitening board and plug our cable on it.

- tried checking the digital signal but we somehow couldn't configure DAQ setting, So actually we couldn't make sure which channel corresponds to our signal.

We turned off many excessive violin mode bandstop filters in the LSC.

Due to some feedforward work by Jenne or EQ some years ago, we have had ~10 violin notches on in the LSC between the output matrix and the outputs to the SUS.

They were eating phase, computation time, and making ~3 dB gain peaking in places where we can't afford it. I have turned them off and Gautam SDF safed it.

Offensive BS shown in brown and cooler BS shown in blue.

To rotate the DTT landscape plot to not be sideways, use this command (note that the string is 1east, not least):
pdftk in.pdf cat 1east output out.pdf

The clue was that the loop X arm POX loop looked to have low (<3dB)) gain margin around 600 Hz (and again at 700 Hz). Attachment #1 shows a comparison of the OLTF for this loop (measured using the IN1/IN2 method) before and after our change. We hypothesize that one of the violin filters that were turned off had non-unity DC gain, because I had to lower the loop gain by 20% after these turn-offs to have the same UGF. I updated the snap files called by the arm locking scripts.

I think I caught all the places where the FM settings are saved, but some locking scripts may still try and turn on some of these filters, so let's keep an eye on it.

Friday, we were seeing a 2 Hz harmonic series in all of the PEM channels. Today I found that some bad person had put in a 4V (!) signal into one of the channels with a signal generator. The generator was also sneakily stuck way back inside the DCU rack. NO SECRET SIGNAL INJECTIONS!

Since the ADC has a 2Vpk range, this was saturating and putting in harmonics in all the adjacent channels. I disconnected it and turned off the function generator.

I changed the office area thermostate near Steve's desk from 68F to 73F today. Please do not change it.

If anyone from facilities comes to adjust something, please put the details in the elog on the same day so that we can know to undo that change rather than chase down other drifts in the system.

Not sure what's wrong, but the workstation desk is freezing cold again and the room temp is 18degC (64degF).

I've compared offline Wiener filtering with online static + adaptive filtering for MC_F with GUR1_XYZ and GUR2_XYZ as witness signals

Note: online filter works up to 32 Hz (AI filter at 32 Hz is used). There is no subtraction up from this frequency, just MC_F was measured in different times for online and offline filtering. This difference in MC_F in frequency range 20-100 Hz showed up again as before with microphone testing.  One can see it in 1 minute. Smth is noisy.

Reasons why online filter is worse then offline:

1. FIR -> IIR conversion produces error. Now I'm using VECTFIT approximation with 16 poles (splitting into 2 filter banks), this not enough. I tried to use 50 and split them into 5 filter banks, but this scheme is not working: zpk -> sos conversion produces error and the result filter works completely wrong.

2. Actuator TF. VECTFIT works very good here - we have only 1 resonance. However, it should be measured precisely.

3. Account for AA and AI filters that rotate the phase at 1-10 Hz by ~ 10 degrees.

[Offsets in MC]

 I have introduce an offset in MC2 PIT because the PZT1 again started railing.

Right now the PZT1 EPICS value is within the range happily.

Please keep this MC eigen axis as a nominal configuration.

[IFO-configure script]

 I have modified the IFO configure scripts such that XARM and YARM are locked with POX11 and POY11 respectively.

A big advantage in use of POX and POY is that we don't need to misalign ITMs when we align each arm.

Those scripts are now available from the C1IFO_CONFIGURE screen as usual.

At 1Y2 rack, I measured offset voltage of the common mode servo (D040180-B) with the gain of it varied.

For now, all signal cables that come into or go out of the common mode servo are not plugged.

I will upload the data I took and report the result later.

I report the results of the measurement to know how offset voltage of common mode servo changes when the gain is changed.

- Motivation: If discontinuous change of the offset happens when we change the gain, it could cause saturation somewhere and so make the length control down. So, we want to estimate effect of such discontinuous change.

- Method: In 1 (or In 2) was terminated with 50 ohm, and the output voltage at Out 2 was measured with a multimeter (D040180-B).

- Results are shown below. Acquired data are attached in .zip.

The upper shows input equiv. offset. The lower shows offset measured at Out 2.

As for both In1 and In2, strange behaviors can be seen between -17 dB and -16 dB.

This is because 5 amplifiers (or attenuators) are simultaneously enabled/disabled here. Similar situation occurs every change of 8 dB gain. 

Eric, Den

We noticed that PRMI RIN is high when the cavity is locked on RELF33 I&Q signals. We compared the level of power fluctuations when PRMI was locked using REFL11, REFL55 and REFL 33. Attached plot "prmi_rin" shows the spectra.

Then we excited PRM and measured length to RIN coupling when PRMI was locked on REFL33 I&Q. DC offset of PRCL is only 3%. But MICH offset seems to be ~nm. When we gave offset of -15 cnts to the servo, power fluctuations improved by a factor of few.

Then we looked at DRMI. It seems that SRC macroscopic length is off but we still could lock it stably. To account for macroscopic length detuning we had to rotate REFL55 phase from 25 degrees to 50 degrees. Power at AS port increased by factor of ~2 compared to PRMI configuration. SPOP18 is decreased only by 30%. Attached plot "drmi_power" shows POP18, POP90, POPDC and ASDC in PRMI and DRMI configurations.

Plot "src_ol" shows srcl OL transfer function. UGF is 70 Hz. We have also centered SRM OL and copied the servo filters from PRM, gains are set to keep UGF at ~0.1 Hz and 7 Hz

This is a more detailed procedure:

1. Phase: REFL11: 19 degree, REFL55: 50 degrees (25 degrees for PRMI configuration)

2. Input matrix:

3. Servo parameters:

PRCL: gain = -0.02, FM4,5 + trigger FM2,3,6,9

MICH: gain = 0.8, FM4,5 + trigger FM2

SRCL: gain = 0.25, FM4,5 + trigger FM2,3

4. Triggering:

signal is SPOP22 , levels 40:25

I added lubricating oil to roughing pumps RP1 and RP3 yesterday and this morning.  Also, I found a nearly full 5 gallon jug of grade 19 oil in the lab.  This should set us up for quite a while.  If you need to add oil the the roughing pumps, use the oil in the quart bottle in the flammables cabinet.  It is labeled as Leybold HE-175 Vacuum Pump Oil.  This bottle is small enough to fill the pumps in close quarters.

 Small oil drops were found during prevent inspection of the vertex crane. They were wiped off. It took 231 days to grow this size.

16 years old Lightwave NPRO M126-1064-700, sn 415 power output is tripping continously to zero.

The Lightwave Controller 125/126-OPN-POS sn516 was used in this test. Settings were lowered to close to nominal values without any success.

One can not determine what is broken: head or controller. This NPRO head was under Manasa's desk.
 

Rod Luna picked up these computers for Larry Wallace yesterday: Dell Inspiron 530, Dell Dimension 4600 and SunBlade 1000

I've restarted the old daqd on fb until I can figure out what's going on with the symmetricom driver on fb1.

Steve:  Jamie with hair.... long time ago
 

Gautam and Steve,

The "called 225 lbs" steel crane load measured right on 102 kg

The trick to the measurment to maintain 1 mm gap to the central cilynder of the load cell.

The lead plate stabilized the large load.

gautam: some additional notes:

1. the wiring on the Omega controller unit as given to us was wrong - I had to fix this on the D-sub connector in order to get the load cell to work. something to check for the other units.
2. the main difficulty in doing this calibration run was that the readback is very sensitive to tilts of the load relative to the sensor.
3. the problem is complicated by the fact that the load cell itself does not have a flat surface - it has a ring that protrudes above the flat face of the cylindrical load cell by a few mm as Steve mentioned.
4. so in order to measure the weight of our stacks, we have to mitigate this problem and ensure that the full load of the stack is normally incident on the load cell - if the load cell itself is somehow torqued during the measurement because of the distribution of the load on it being uneven, we get an inaccurate measurement.
5. In this calibration measurement, we think the error is <1% (true mass is 102kg, we measure 104kg on the meter which seems reasonable as the sum of the donut + lead plate)

I am at the PSL table taking what is hopefully the last set of pictures for the diagram. Woohoo.

 I'm out, HEPAs are back at 20%.

For a last few days I've been working on oaf and simulink model to simulate it. First I did online subtraction from MC when MC_L path was enabled. Inside my code I've added a sum of squares of filter coefficients so we can monitor convergence of the filter.

To to this I've measured path from OAF output to input without AA and AI filters. Then made a vectfit using 2 poles and zeros. Foton command

zpk( [-2.491928e+03;5.650511e-02], [-4.979872e+01;-3.278776e+00], 6.011323e+00)

My simulink model consists of 3 parts:

• cavity with seismic noise at low frequencies, 1/f^2 noise at medium frequencies and white noise at high frequencies
• this cavity is locked using feedback compensation filters that we use to lock arms
• locked cavity with adaptive filter

Adaptive filter in the model uses online c-code. It is connected to simulink block through an S-function. Sampling frequency of the model is 10 kHz. It works fairly fast - 1 sec of simulation time is computed in 1 sec.

I've tested FxLMS algorithm and MFxLMS algorithm that is faster. I plan to test 2 iir adaptive algorithms that are already coded.

 Excitation generator created these harmonics. When I applied low-pass butterworth filter, I've got the result of online filtering close to simulations. On the second graph blue is biquad filter output spectrum, red corresponds to DF2. 1 Hz sin wave was filtered with a notch filter of Q=100, depth=300 at 1 Hz.

I tried to filter MC_F from seismic noise measured by GUR1 seismometer. I've used 8000 tap filter, downsample ratio=8, delay=1. In the Figure the output of the filter is presented with MC_F signal.

We can see that output is close to the MC_F, but the phase for some reason is not zero. It should not be at 1 Hz - 10 Hz due to the actuator. But below these frequencies I do not see any reasons for the output phase to differ from MC_F phase. But it is possible, the phase of the actuator is evaluated very rough and the adaptive filter can't match it.

The 2W PSL laser is turned off.  The danger laser lights are not illuminated at the entry doors because of malfunctioning electronic circuit!!!

Laser safety glasses are still required!  Other lasers are in operation!

I had forgotten to fix the DNS setup on op340m and so our slow, perl, PID loops for the laser were not running well (and that's why the FSS-FAST has been saturating).

I edited /etc/resolv.conf on there and then did /usr/sbin/reboot as super-user.

op340m:~>more /etc/resolv.conf
domain martian
nameserver 192.168.113.104
nameserver 131.215.125.1
nameserver 131.215.139.100
nameserver 131.215.9.49
op340m:~>date
Fri Jun 20 19:06:37 PDT 2014

The FSSSlowServo.pl now seems to be holding the NPRO PZT to ~6 V. I twiddled the PID settings a little bit to make sure nothing was squirrelly. Seems OK. Time constant of the loop is ~1 minute.

As a reminder, op340m runs our autoburt for all the FE machines, VME IOCs, does the watchdog threshold rampdown, and also the RefCav and NPRO temperature control.

 We had fiddled with the scripts/general/scripto_cron script to try and get the MC auto locker working on ottavia, but in doing so broke op340m's reliance on it to run it's cron jobs, like FSSSlowServo.

I've reverted scripto_cron to its original state, and the FSS slow servo starts up again.  

However, scripts like this that we want to always have on seem to be a better fit, to me, for the init system, like we do with daqd and nds on the FB. op340m's inittab looks different than what I'm used to, so I'm not making any changes; this is just a thought. 

MC autolocker is still being ran from an Ottavia GUI terminal; I'll try to get it consistently running on megatron, as suggested in  ELOG10039, now that caget/caput issues seem to be sorted. 

Addendum: I've changed the MC auto locker script to have megatron as its host. Haven't yet gotten it to run automatically; it's running in a detached tmux terminal. I'll finish it up tomorrow. 

I modified the perl script which does our hourly autoburt so that it can run on megatron instead of op340m (old Solaris machine). Nothing major, just some path stuff. That was the last function of op340m that I know of, so after a week of watching this we ought to be able to power it off and send it to e-waste.

Seems to work so far. It complains about some models that aren't running but mostly it reports successful snapshot taking based on the .req files.

Unfortunately, it seems that its only doing the new target directory, so its missing all of our old VME machines which still use the /cvs/cds/caltech/target area.

But I think Gautam and Jamie and Aidan have volunteered to start our slow controls upgrade by moving the EX slow controls to Acromag and into the new target area. We ought to modify the CRON to point at the old directory for now, but its a temporary fix hopefully.

I added op540m's display 0 (the northern-most monitor in the control room) to the MEDM screens webpage: https://nodus.ligo.caltech.edu:30889/medm/screenshot.html

Now we can see the StripTool displays that are usually parked on that screen.

I think op540m has finally bitten the dust.  I noticed that both of its screens were black, so I assumed that it had crashed due to known graphics card issues or something.  But upon closer inspection, it is way more dead than that.  I checked that it does have power (at least the power cable is securely plugged in at both ends, and the power strip its on is successfully powering several other computers), but I can't make any lights or anything come on by pressing the power button on the front of the computer tower.

Immediate consequences of op540 not being operational are the lack of DMT, and the lack of Alarms. 

Joe is doing an autopsy right now to see if its really dead, or only 'mostly dead'.

EDIT: Joe says maybe it's the power supply for the computer.  But he can't turn it on either.

This is a reminder (mainly for Steve, who somehow doesn't believe these things) that op540m is not to be used for your general pleasure.

No web, no dataviewer, no DTT. Using these things often makes the graphical X-Windows crash. I have had to restart the StripTool, our seismic BLRMS and our Alarms many times because someone uses op540m, makes it crash, and then does not restart the processes.

Stop breaking op540m, Steve!

I turned the StripTool and ALARMS and BLRMS back on on op540m. Looks like it has been rebooted 5 days ago and no one turned these back on. Also, there was a bunch of junk strewn around its keyboard which I restrained myself from throwing in the trash.

The BLRMS trends should be active now.

The squeezing open frame rack was moved from the south side of the PSL enclosure to the north side of the SP table.

AC power breaker is PC-2  #1

I built a Simulink model of the magnetic levitation system and try to explain the dip in the open-loop transfer function that was observed.

One can download the model in the svn. The corresponding block diagram is shown by the figure below.

Here "Magnet" is equal to inverse of the magnet mass. Integrator "1/s" gives the velocity of the magnet. A further integrator gives the displacement of the magnet.

Different from the free-mass response, the response of the magnet is modified due to the existence of the Eddy-current damping  and negative spring in the vertical

direction, as indicated by the feedback loops after two integrals respectively. The motion of the magnet will change the magnetic field strength which in turn will pick

up by the Hall-effect sensor. Unlike the usual case, here the Hall sensor also picks up the magnetic field created by the coil as indicated by the loop below the mechanical

part. This is actually the origin of the dip in the open-loop transfer function. In the figure below, we show the open-loop transfer function and its phase contributed by both

the mechanical motion of the magnet and the Hall sensor with the black curve "Total". The contribution from the mechanical motion alone is shown by the magenta curve

"Mech" which is obtained by disconnecting the Hall sensor loop (I rescale the total gain to fit the measurement data due to uncertainties in those gains indicated in the figure). 

The contribution from the Hall sensor alone is indicated by the blue curve "Hall" which  is obtained by disconnecting the mechanical motion loop. Those two contributions

have the different sign as shown by the phase plot, and they destructively interfere with each other and create the dip in the open-loop transfer function.

In the following figure, we show the close-loop response function of the mechanical motion of the magnet.

As we can see, even though the entire close loop of the circuit is stable, the mechanical motion is unstable around 10 Hz. This simply comes from the fact that

around this frequency, the Hall sensor almost has no response to the mechanical motion due to destructive interference as mentioned.

In the future, we will replace the Hall sensor with an optical one to get rid of this undesired destructive interference.

We've had a devil of a time getting V1 to open, due to the Interlock code. 

The short story is that if C1:Vac-PTP1_pressure > 1.0, the interlock code won't let you push the button to open V1 (but it won't close V1). 

PTP1 is broken, so the interlock was frustrating us.   It's been broken for a while, but this hasn't bitten us till now.

We tried swapping out the controller for PTP1 with one of Bob's from the Bake lab, but it didn't work. 

It said "NO COMM" in the C1:Vac-PTP1_status, so I figured it wouldn't update if we just used tdswrite to change C1:Vac-PTP1_pressure to 0.0.  This actually worked, and V1 is now open.  This is a temporary fix.

The swapped in 307 convectron gauge controller  is very likely to have the  RS232 connection  wired differently from the old one.

PRP gauge has now the same error message as the PTP1:  "no comm"  I would look at RS232 wiring of the PRP gauge on the broken

controller and adapt the swapped in one to communicate. The PRP was reading 620 Torr before the swap.

Back when Gabriele was here, he and I implemented online calibration of the oplevs, into microradians.  A consequence of this is that all of the XY-plots on the medm screens were too small. 

I have gone through all the screens that I could think of (SUS_SINGLE, SUS_SINGLE_OPTLEV_SERVO, OPLEV_MASTER, OPLEV_SUMMARY, OPLEV_SUMMARY_SMALL_SCALE, IFO_OVERVIEW) and made the range of the XY-plots +/- 100, rather than the old scale of +/-1. 

I have also added red boxes behind the numbers on many (but not yet all) of these screens, so that you can see when (a) the oplev sum is too low, or (b) either the pit or yaw value is over 50 microradians. 

While I was putzing around on the IFO overview screen, I also made the oplev sum numbers clickable, with the related display being that optic's oplev servo screen.

The ETMX oplev signal looks kind of dead compared to the ETMY. It has no features in the spectra and the SUM is pretty low.

I noticed that the cal fields are still set to 1. To get it close to something reasonable, I calibrated it vs. the SUSPIT and SUSYAW values by giving it a step in angle and using 'tdsavg' plus some arithmetic.

OLPIT =  45 urads/ count

OLYAW = 85 urads / count

These are very rough. I don't even know what the accuracy is on the OSEM based calibration, so this ought to be redone in the way that Jenne and Gabriele did before.

The attached image shows the situation after "calibration" of ETMX. This OL system needs some noise investigation.

As another proof that sometime is ill with ETMX Optical Lever:

We scanned the ETMX bias in PIT using ezcastep and saw that the OL response is very screwy. In the attached, you can see that the ETMX SUSPIT signal shows that the actual motion is good and linear. In fact, our sus diagonalization is extermely good and there's almost no signal in SUSYAW.

I used the same OSEM SUSPIT/YAW method as before to calibrate the ETMY optical lever signals. They were off by a factor of ~10.

ETMY Pitch     300  /  26    (old/new)     urad/counts

[Jenne, Gabriele]

Optical lever calibrations:

ITMX pit calibration = -9.07 cts/mrad

ITMX yaw calibration = -12.33 cts/mrad

ITMX plot:

BS pit calibration = -22.86 cts/mrad

BS yaw calibration = -24.14 cts/mrad

BS plot:

Method:  Similar to Manasa and Yuta's method last month.  We mounted each oplev QPD on a micrometer translation stage, centered the beam using the steering mirror, then used tdsavg to get 10 second averages of the _INMON channel for various settings of the micrometer stage.  For BS, we had to take out the PRM oplev to make room for the translation stage.  All QPDs were remounted in their original positions, within less than 1mm.  Measured the out-of-vac distances with the laser disto-meter, and the invac distances from the optic to the window from the CAD drawing.

Copying from other elog entries,

elog 8232:

We calibrated oplev for ITMY. Calibration factor for C1:SUS-ITMY_OL(PIT|YAW)_IN1 are;
  OLPIT: 6.29 +/- 0.11 counts/mrad
  OLYAW: 5.74 +/- 0.09 counts/mrad

elog 8221:

We calibrated oplev for PRM. Calibration factor for C1:SUS-PRM_OL(PIT|YAW)_IN1 are;
  OLPIT: 15.6 +/- 0.3 counts/mrad
  OLYAW: 17.8 +/- 0.3 counts/mrad