[Jim Ryan]

The PEM model has been modified now to include a block called 'JIMS' for the JIMS(Joint Information Management System) channel processing. Additionally I added test points inside the BLRMS blocks that are there. These test points are connected to the output of the sqrt function for each band. I needed this for debugging purposes and it was something Jenny had requested.

The outputs are taken out of the RMS block and muxed, then demuxed just outside the JIMS block. I was unable to get the model to work properly with the muxed channel traveling up or down levels for this. Inside the JIMS block the information goes into blocks for the corresponding seismometer channel.

For each seismometer channel the five bands are processed by comparing to a threshold value to give a boolean with 1 being good (BLRMS below threshold) and 0 being bad (BLRMS above threshold). The boolean streams are then split into a persistent stream and a non-persistent stream. The persistent stream is processed by a new library block that I created (called persist) which holds the value at 0 for a number of time steps equal to an EPICS variable setting from the time the boolean first drops to zero. The persist allows excursions shorter than the timestep of a downsampled timeseries to be seen reliably.

The EPICS variables for the thresholds are of the form (in order of increasing frequency):

C1:PEM-JIMS_GUR1X_THRES1

C1:PEM-JIMS_GUR1X_THRES2

etc.

The EPICS variables for the persist step size are of the form:

C1:PEM-JIMS_GUR1X_PERSIST

C1:PEM-JIMS_GUR1Y_PERSIST

etc.

I have set all of the persist values to 2048 (1 sec.) for now. The threshold values are currently 200,140,300,485,340 for the GUR1X bands and 170,105,185,440,430 for the GUR1Y bands.

The values were set using ezcawrite. There is no MEDM screen for this yet.

PEM model was restarted at approx. 11:30 Mar. 7 2012 PST.

The RF separator installed comprises of the Minicircuits filters cascaded as in the figure below.
This has one input and 4 output ports for 11, 22, 30-60, and 110MHz signal.
As seen in this entry #6167, we have 22 and 110MHz signals together with 11, 44, 66MHz signals.
They may be demodulated via a harmonic characteristic of the mixers. (Remeber mixers are not multipliers.)

 Of course the big concern is the impedance matching for those signals as usual.
The 2nd attachment shows measured impedance of the circuits with all of the ports terminated.
From the complex impedance, we can calculate the reflection coefficient. The 44 and 110MHz
components look correctly matched while the others seems largely reflected.
This certainly is not a nice situation, as the reflection can make the amplifier next to the PD unhappy
(although the reflected power is tiny in our case).

In our case more eminent problem is that the amplitude of the 22MHz signal can vary depending on the cable length by
factor of 10 in amplitude. (c.f. VSWR on the 2nd attachment.)

The transmission to each port was measured. The separation of the signals looks good. But the attenuation of the
targetted signals (i.e. insertion losses) are qulitatively consistent with the impedance. Again these losses are depend
on the cable length.

 I swap an OSA at PSL and OSA at REFL. It was because the PSL-OSA had a better resolution, so we place this better one at REFL. The ND filter (ND3) which was on the way to REFL OSA was replaced by two BSs, because it was producing dirty multiple spots after transmitting.

 This is the calibrated MICH noise budget on Mar 5. There was a sharp peak at 1Hz and a blob on 3 Hz. The demod phase was adjusted for AS55Q.

Just a quick report on the REFL OSA.

The attached plot below shows the raw signal from the REFL OSA which Keiko installed in this afternoon.

When the data was taken the beam on the REFL OSA was a direct reflection from PRM with the rest of the suspended mirrors misaligned.

One of the upper and lower 11 MHz sidebands is resolved (it is shown at 0.12 sec in the plot) while the other one is still covered by the carrier tail.

The 55 MHz upper and lower sidebands are well resolved (they are at 0.06 and 0.2 sec in the plot).

One of the oscilloscopes monitoring the OSA signals in the control room has a USB interface so that we can record the data into a USB flash memory and plot it like this.

I'm puzzled why the 11MHz peak can be such high considering 1.7~2 times smaller the modulation depth.

This is the MICH noise budget on 6th March. 1Hz peak got a bit better as the BS sus control gain was increased.

 I noticed that NDS2 was not running on mafalda as it should be. Instead, there were a couple of zombie MEDMs using up 99% of the CPU. I killed the zombies and have run the 'build channel list' script. When it finished, I tried to restart the nds server, but got the following error in the log file. Email has been dispatched to JZ.

mafalda:logs>less nds2-mafalda-201203072111.log

Configuring from file: nds2.conf
Allow list: ALL
terminate called after throwing an instance of 'std::runtime_error'
  what():  Insufficient arguments

I was also wondering about the same thing, comparing with what Mirko obtained before with the same OSA ( #5519).

 kiwamu, keiko

We measure the REFL OSA spectrum when (1) direct reflection from the PRM (2) CR lock at PRC (3) SB lock at PRC. When CR lock, both SBs are reflected from the PRC and when SB lock (ref line), some SB is sucked by PRM and looked lower than the other two lines.

Here is the recent two noise budgets of MICH, with the old measurement by Jenne. The most latest Mar 6 data is quite close to the old data, even better around 20-30 Hz. Probably some scattering source was improved?

[Keiko / Kiwamu]

 Some updates on the locking activity:

• Started summarizing the data of the Michelson lock in a wiki page:
  • https://wiki-40m.ligo.caltech.edu/Interferometer_Characterization/Michelson_Noise_Budget
  • Some careful analysis of the OSA signals are missing
    • Besides the analysis we need to work on the mode matching to get a better resolution as Mirko achieved.
• Gradually moving on to the PRMI lock
  • The lock stays for reasonably a long time (~20 min or more)
  • POP22/110 demod signals seemed just ADC noise.
  • A first noise budget is in process
    • The glitches make the noise level worse above 40 Hz or so in both the MICH and PRCL budgets.
  • Sensing matrix will be measured tomorrow
  • The data will be also summarized in a wiki page

I disabled the feedback to the PZT1 PITCH in the Y arm dithering scripts so that it won't push the beam away from the good point.

Currently one has to do a manual alignment only for the PZT PITCH but the rest of DOFs are still able to be automatically aligned with the script.

Optical spectrum analyzers like the Attachment made by Coherent , Meles Griot- CVI and Spectral Product are all discontinued.

The 40m have Coherent models C240 analyzer with controller C251 Their Finesse measured in 2004: sn205408  F302,  sn205409  F396,

Jenne borrowed Jan's Meles Griot model 13SAE006, Peter King has the same model. FSR 300 MHZ, finnees 200 minimum

We tried to measure the sensing matrix for MICH and PRCL last night. They look too much mixed as we expect... the matrix may be posted later. We suspect the IX and IY of the MICH excitation is not balanced very well, although Kiwamu adjusted that about two weeks ago, and it is mixing the dof. We'll try to balance it again, ans see the matrix. 

Keiko, Kiwamu

Over-sized local laser emergency switch was held by large C clamp at the south end. This was replaced by a smaller one and it is mounted with magnets.

The Innolight laser was turned off, while the interlock was wired.

DAQ reload/restart was performed at about 1315 PST today. The previous ini file was backed up as c1pem20120309.ini in the /chans/daq/working_backups/ directory.

I set the following to record:

The two JIMS channels at 2048:
[C1:PEM-JIMS_CH1_DQ] Persistent version of JIMS channel. When bit drops to zero indicating something bad (BLRMS threshold exceeded) happens the bit stays at zero  for >= the value of the persist EPICS variable.
[C1:PEM-JIMS_CH2_DQ] Non-persistent version of JIMS channel.

And all of the BLRMS channels at 256:
Names are of the form:
[C1:PEM-RMS_ACC1_F0p1_0p3_DQ]
[C1:PEM-RMS_ACC1_F0p3_1_DQ]

On monday I intend to look at the weekend seismic data to establish thresholds on the JIMS channels.

256 was the lowest rate possible according to the RCG manual. The JIMS channels are recorded at 2048 because I couldn't figure out how to disable the decimation filter. I will look into this further.

ITMX and ITMY balance for the MICH excitation (lockin) is adjusted again. Now it's ITMx = -0.992, ITMy = 1 for MICH (lockin output matrix values).

RA: what were the old values? Does this change make any difference for the signal mixing noticed before?

 I calculated the DRMI RAM LSC matrix with RAM and the operation point offsets.

• configuration: C1 DRMI
• RAM is added by an Mach-Zehnder ifo placed before the PRM
• demodulation phases are optimised for each DoF
• the operation points offset from the PDH signals are calculated and added to the optical configuration as mirror position offsets
• Then the matrix is calculated with the offsets and the RAM
• The set of the scrips are found as RAMmatrix.m, normMAT.m, newGetMAT.m,  on CVS/ifomodeling/40m/fullIFO_Optickle. They are a bit messy scripts at this moment.

Results:

(1) No RAM LSC matrix

(2) With 1% RAM mod index of PM (normalised by (1) )

(3) With 5% RAM mod index of PM (normalised by (1) )

Adding some more results with more realistic RAM level assumption.

(4) With 0.1% RAM mod index of PM (normalized by (1) )

(5) With 0.5% RAM mod index of  PM (normalized by (1) )

I have completed the calibration of the demod board efficiencies.  Here is the schematic of the set-up.

The data is given below and the data-file is attached in several different formats.

The punch line is -- the sensing matrix still looks strange in the PRMI configuration.

I have been measuring the sensing matrix of the PRMI configuration because it didn't make sense (#6283).

One strange thing I have noticed before was that all the I-phase signals showed a weird behavior -- they fluctuate too much in time series.

Tonight I measured the sensing matrix again but this time I recorded them as a function of time using the realtime LOCKINs in the LSC front end.

The attached plots are the responses (optical gains) of PRCL and MICH in watts / meter at various sensors in time series.

I will explain some more details about how I measured and calibrated the data in another elog entry.

Here I describe the measurement of the sensing matrix.

Motivations

  There were two reasons why I have been measuring the sensing matrix :

1.  I wanted to know how much each element in the sensing matrix drifted as a function of time because the sensing matrix didn't agree with what Optickle predicted (#6283).
2.  I needed to estimate the MICH responses in the 3f demodulated signals, so that I can decide which 3f signal I should use for holding MICH.

 I will report #2 later because it needs another careful noise estimation.

Measurement

 In order to measure the sensing matrix, the basic steps are something like this:

1. Excite one of the DOF at a certain frequency, where a notch filter is applied in the LSC servos so that the servos won't suppress the excitation signal.
2. Demodulate the LSC signals (e.g. C1:LSC-REFL11_I_ERR and etc.,) by the realtime LOCKINs (#6152) at the same frequency.
3. Calibrate the obtained LOCKIN outputs to watts/meter.

LOCKIN detection

• The LOCKIN oscillator excites ITMs differentially
  • In order to purely excites the MICH DOF, the actuation coefficients were precisely adjusted (#6398).
  • Currently ITMY has a gain of 1, and ITMX has a gain of -0.992 for the pure MICH excitation. Those numbers were put in the output matrix of the LOCKIN oscillator.
• The demodulation phase of the LOCKIN system was adjusted to be -22 deg at the digital phase rotator.
  • This number maximizes the in-phase signals while the quadrature-phase signals give almost zero.
  • This number was adjusted when the simple MICH configuration was applied.
• In the demodulations, the LO signals have amplitude of 100 counts to just make the demodulated signals readable numbers.

Calibration of the LOCKINs

Calibration of the responses to watts/meter

Settings and parameters

•  LSC RF demodulation phases
  
  •  AS55 = 17.05 deg (minimizing the PRCL sensitivity in the Q-phase)
  •  REFL11 = -41.05 deg (maximizing the PRCL sensitivity in the I-phase)
  • REFL33 = -25.85 deg (maximizing the PRCL sensitivity in the I-phase)
  • REFL55 = 4 deg (maximizing the PRCL sensitivity in the I-phase)
  • REFL165 = 39 deg (random number)
•  Whitening filters
  • AS55 = 30 dB
  • REFL11 = 0 dB
  • REFL33 = 42 dB
  • REFL55 = 30 dB
  • REFL165 = 45 dB
• MICH servo
  • AS55_Q for the sensor
  • G = -5 in the digital gain
  • FM2, FM3, FM5 and FM9 actiavted
  • UGF ~ 100 Hz
  • Feedback to ITMs differentially
• PRCL servo
  • REFL33_I for the sensor
  • G = 1 in the digital gain
  • FM2, FM3, FM4, FM5 and FM9 activated
  • UGF ~ 100 Hz
  • Feedback to PRM

Next steps:

• Compare the obtained sensing matrix with an Optickle model. Particularly I am interested in the absolute strengths in watts/meter
• Noise estimation of the REFL33_Q as a MICH sensor to see if this sensor is usable for holding MICH.

A feasibility study of the REFL33Q as a MICH sensor was coarsely performed from the point view of the noise performance.

The answer is that :

  the REFL33Q can be BARELY used as a MICH sensor in the PRMI configuration, but the noise level will be at only sub-nano meter level.

  Tonight I will try to use the REFL33Q to control the MICH DOF to see what happens.

(Background)

(Noise analysis)

  I did a coarse noise analysis for the REFL33Q signal as shown in the attached plot below while making some assumptions as follows.

•  Optical gain for MICH = 0.8  W/m (#6403)
  • In the plot below, I plotted a unsuppressed MICH motion which had been measured the other day with a different sensor. This is for a comparison.
•  Shot noise due to DC light on the REFL33 photo diode
  •  With a power of 5.0 mW (#6355)
  • Assume that the responsivity is 0.75 A/W, this DC light creates the shot noise in the photo current at a level of 35 pA/sqrtHz.
  • Then I estimated the contribution of this shot noise in terms of the MICH displacement by calibrating the number with the optical gain and responsivity.
  • It is estimated to be at 60 pm/sqrtHz
• Dark current
  • I assumed that the dark current is 0.52 mA. (see the wiki)
  • In the same manner as that for the shot noise, the dark current is estimated to be at 20 pm/sqrtHz in terms of the displacement
• Whitening filter input referred noise
  • I assumed that it is flat with a level of 54 nV/sqrtHz based on a rough measurement by looking at the spectrum of the LSC input signals.
  • The contribution was estimated by applying some gain corrections from the conversion efficiency of the demod board, transimpedance gain, responsivity and the optical gain.
  • This noise is currently the limiting factor over a frequency range from DC to 1 kHz.
• ADC noise
  • I did the same thing as that for the whitening filter noise.
  • I assumed the noise level is at 6 uV/sqrtHz and it is flat (I know this not true particularly at mHz region the noise becomes bigger by some factors)
  • Then I applied the transfer function of the whitening filter to roll off the noise above 15 Hz.

(Some thoughts)

•   Obviously the limiting noises are that of ADC and the whitening filter.
  • These noise can be easily mitigated by installing an RF amplifier to amplify the RF signals from the REFL33Q RFPD.
  • Therefore this is not the real issue
• The real issue is that the shot noise is already at a level of 60 pm/sqrtHz, and we can't suppress the MICH motion less than that.
  • In order to decrease it, one possibility is to increase the modulation depth. But it is already at the maximum.
  • If the REFL165 RFPD is healthy, it is supposed to give us a bigger MICH signal. But it didn't look healthy ... (#6403)

[Rana / Kiwamu]

 We put some elements in the BS output matrix to mitigate the actuator coupling from SIDE to POS.

As a result the degree of the coupling reduced by a factor of 2 or so.

Rana did the "Q of 5" test on the SIDE damping servo after putting the elements and set the gain to be 40.

The attached screen shot is the new elements that we put in the suspension output matrix.

(How to)

• Excite the SIDE motion by AWG at 3 Hz.
• Monitor the POS signal in DTT
• Try some numbers in the matrix elements until the peak at 3 Hz in the POS signal is minimized

As the PZT1 has not been functional, I have been aligning the Y arm to the input beam instead of aligning the beam to the Y arm.

It turned out that this procedure leads to two extra works everytime after alignments of the Y arm:

1. The Y green beam must be always aligned to the Y arm
   • The amount of the misalignment was found to be relatively big compared with how it used to be.
2. The PSL beat note setup must be always realigned because the Y green path is determined by the orientation of the Y arm.
   • In the past I didn't often realign the beat note path, but currently it needs to be pay more attentions.

Sad ..

 I tried the REFL33Q for controlling MICH in the PRMI configuration (#6407)

The result was --

 It was barely able to lock MICH in a short moment but didn't stay locked for more than 10 sec. Not good.

(Tricks)

The following tasks need to be done in the daytime tomorrow.

• Hook up the DC output of the Y green BBPD on the PSL table to an ADC channel (Jamie / Steve)
• Install fancy suspension matrices on PRM and ITMX [#6365] (Jenne)
• Check if the REFL165 RFPD is healthy or not (Suresh / Koji)
  • According to a simulation the REFL165 demod signal should show similar amount of the signal to that of REFL33.
  • But right now it is showing super tiny signals [#6403]

Dichroic mirror quotes are in the wiki.

ATF is pricy.

We got a good price from Laseroptik, but the wedges are 5 arcminutes. The fused silica grade is 0F, meaning the  homogeneity is 5 ppm instead of 1ppm.  I requested an other large wedge quote on the substrates.We may have to get substrates from somebody else and ship it to Germany

MLT quote is outrageously high

REO is not interested in this low volume job.

A correction on the previous elog about the REFL33Q noise:

 For ITMX, I used the values from the conlog:

2011/08/12,20:10:12 utc 'C1:SUS[-_]ITMX[-_]INMATRIX'

These are the latest values in the conlog that aren't the basic matricies.  Even though we did a round of diagonalization in Sept, and the 
matricies are saved in a .mat file, it doesn't look like we used the ITMX matrix from that time.

For PRM, I used the matricies that were saved in InputMatricies_16Sept2011.mat, in the peakFit folder, since I couldn't find anything in the Conlog other than the basic matricies.

UPDATE:  I didn't actually count the number of oscillations until the optics were damped, so I don't have an actual number for the Q, but I feel good about the damping, after having kicked POS of both ITMX and PRM and watching the sensors.

In the last post, I showed that SRCL element in the MICH sensor (AS55I-mich) is chaned 1% due to RAM.

Here I calculated how is this 1% residual in MICH sensor (AS55 I-mich) shown in MICH sensitivity. The senario is:

(1) we assume we are canceling SRCL in MICH by feed forward first (original matrix (2,3) element).

(2) SRCL in MICH (matrix(2,3) is changed 1% due to RAM, but you keep the same feed forward with the same feedforward gain

(3) You get 1% SRCL residual motion in MICH sensor. This motion depends on how SRCL is quiet/loud. The assumed level is

Pollution level = SRCL shot noise level in SRCL sensor  x  SRCL closed loop TF  x  1% residual .... the following plot.

AS sensor = AS55I-mich  --- SN level 2.4e-11 W/rtHz ------- MICH SN level 6e-17 m/rtHz

SRCL sensor = AS55 I-SRCL --- SN level 2e-11 W/rtHz ---  SRCL SN level 5e-14 m/rtHz

 The REFL165 RF output was not reaching the Demod board.  The RF cable was disconnected.  I fixed that and then I put in a RF signal at 165MHz , 1.66 mVrms at the test input  (100Hz off set from the 165MHz LO) and saw that the 100 Hz demodulated signal was visible in the dataviewer. 

Will complete the Optical RF power -> CDS counts calibration tomorrow morning. 

This is the simulated signals to compare with the original post #6403

Kiwamu and Koji

The target is to realize DRMI or PRMI + one arm with ALS.

The focus of the night is to achive stable lock of the PRMI (SB resonant) with 3f signals.
Particularly, REFL165 is back now, we are aiming to see if any of the 165 signals is useful.

We made a comparison between  REFL33Q/REFL165Q/AS55Q to find any good source of MICH.
However, none of them showed a reasonable shape of the spectra. They don't have reasonable coherence between them.

Nonetheless, we have tried to lock the IFO with those REFL signals. But any of them were useful to keep the PRMI (SB resonant).
The only kind of stable signal for MICH was AS55Q as we could keep the PRMI locked.

These are the measurements for estimating the amplitude of the signal recorded in the CDS when a known amount of modulated light is incident on the photodiode. 

I mounted the PD characterisation setup onto a small breadboard which could then be placed close AP table.  I then placed position markers for REFL165 on the AP table before moving it onto my small breadboard.  The AM laser was driven by an RF function generator (Fluke 6061A) at a frequency of 165.98866 MHz, which is 102 Hz offset from the 165MHz LO.  The power level was set at -45dBm.  This power level was chosen since anything higher would have saturated the AntiAliasing  Whitening Filters.  The counts in the CDS were converted to voltage using the ADC resolution = 20V per 2^16 counts.

 When the 166MHz power is decreased by a factor of 2 the amplitude of 102Hz wave recorded in CDS goes down by sqrt(2) as expected.   The RF AM power incident on the REFL165 was estimated to be 0.011mW(rms)  (case #1 in the above table)  using the DC power ratio and using the transimpedance of the 1611 BBPD to be 700 Ohms.  This produces a 171 mV amplitude wave at 102 Hz.  I then stepped down the power by factor of 2 and repeated the measurement. 

(These numbers however are not agreeing with the power incident on REFL165 if we assume its transimpedance to be 12500.  It will take a bit more effort to make all the numbers agree.  Will try again tomorrow)

Here is a picture of the small black breadboard on which I have put together the PD characterisation setup.  It would be great if we can retain this portable set up as it is, since we keep reusing it every couple of weeks.  It would be convenient if we can fiber couple the path to the PD under test with a 2m long fiber.  Then we will not have to remove the PD from the optical table while testing it.

 This is totally sweet Suresh!  I don't remember how much more fiber is coiled up under the plate that has the "Jenne Laser" label, but there's a reasonable amount.  It's not 2m, but maybe we can just extend the blue snakey thing some?

 To characterize the RF V to counts we need to know the state of the whitening filter board. Was the filter on or off ? What was the value of the whitening gain slider?

The MC alignment servo wasn't great in the last 1 hour or so as it kept disturbing the MC lock. It was found to be due to some offsets in the MC trans QPD signals.

I put some values to cancel the offsets and then the lock became stable.

This is a first aid. So we need to take a closer look at the QPD signals and also probably the spot position on the QPD.

The symptom was that every time the alignment servo was engaged, at the beginning the amount of the transmitted light went to 27000 counts, which is good.

However, then the amount of the transmitted light slowly decreased in a time scale of ~ 20 sec or so, ending up with destruction of the MC lock.

According to the time scale I suspected that the servos using the trans QPD signals were doing something bad because their control width had been designed to be slow and slower than the rest of the servo loops.

I switched off the servos, called C1:IOO-TRANS_PIT and C1:IOO-TRANS_YAW and found the MC stayed locked stably with 27000 counts of the transmitted light.

Leaving the trans QPD servos off, I zeroed the offsets and then switched them on. It worked.

The values below are the current offset that I put.

                C1:IOO-MC2_TRANS_PIT_OFFSET = -0.115203
                C1:IOO-MC2_TRANS_YAW_OFFSET = -0.0323576
 

I've turned off the power of the STS-2 readout box as it provides outputs with ~10 Volts DC offset! AA filter box works in the range -2 +2 Volts, so we do not have any useful information anyway. I'll adjust the mass positions in the seismometer.

 The filter was ON and the whiterning filter gain was 45dB

Today is janitor day. It still does not explain why the 2W Innolight tripped off about an hour ago. All back to normal.

.......................................................I asked Keven later, he admitted hitting the emergency shut off next to the chemical storage cabinet.

Fred Goodbar of Konecrane has completed the annual certified crane inspection and maintenance of our cranes as required in safety document.

They are in good working condition and safe to use.

There was something wrong with the Beam Scan PC.  The  mouse and screen were not responding and the PC was asking for drivers for any new hardware that we plugged in.  We called in the services of Junaid and co. since we do not have a Win98 Second Edition installation disk in the lab.   Junaid came with the disk, we changed the screen and the mouse and installed everything. 

We tried to get the network going on the PC so that we could update stuff easily over the net.  This didnt succeed. For now, we still have to depend on a Win98se CD to get drivers if any new hardware is connected to this machine. 

For future reference, some notes:

1)  We will get a copy of Win98SE for the lab from Junaid

2) We have to use a USB mouse from Dell. We have several spares of this. The drivers for these are present in the machine. 

The Beam Scan is working okay now.  We will proceed with the beam profile measurements.  

ETMY sus damping was disabled. Green locking laser and associated electronics turned off. Computers and power supplies turned off at rack 1Y4

The electricians picking up ac power from 1Y4 manual disconnect box and installing conduit line to ISCT-ETMY east end optical table.

There will be no more daisy chaining this way. 

 The power is back on at ETMY . c1iscey has not been restarted.

Now I'm turning ac power off at ETMX for the same job to be done.

 The power was turned back on at 4pm It took some time for Suresh to restart the computers. We have damping but things are not perfect yet. Auto BURTH did not work well.

 The Laseroptik quote is here.The 2 degrees wedge cost is $40 on each optics!  See wiki