We quickly looked at the damping filters we have on ETMX to compare to the L1 version. Things are pretty similar, other than 60 Hz notches and different gains. In yaw LLO uses a factor of 4 more gain, in length we have a factor of 2 more gain, in pitch we have a factor of 3 more gain.
I wanted to see if our ETMY F2P problems were saturation related. So I looked at the current monitors for the L1 OSEMs.
One thing that was quickly noticed is that while the three others show <100 counts in the readback when there is no drive signal, the LR coil shows -4000 counts. (For scale, saturation occurs around 13000 counts for the three normal looking coils).
This would seem to indicate some problem, possibly removing a third of our range in one direction of one of our coils. However, when you actually drive the coils to saturation, they all seem to have the same range from the point of view of the DAC, just with LR readback offset from zero by 4000 counts.
So this could just be a problem in the readback, not clear without looking at the analog electronics.
J. Kissel, N. Smith-Lefebvre After Nic showed me some of the details of what he found, I saw that a lot of the channels had an offset of not just 4000 [cts], but an infamous value 4300 [ct]. This could very well be an old nasty problem in the analog electronics where one leg of the SCSI connector on the back of the ADC card in the IO chassis shorted. It's best described here: LLO aLOG 1857 Most likely these offsets have been there since these chassis were first cabled up, and never fixed. Other instances where this bug bit us: LHO aLOG 5385 LLO aLOG 1853 I'll add this to Integration Issue 9, which continues to gather dust.
(Alexa, Evan, Sheila)
We made a transfer coeffecient measurement of the M0, L1, L3 stages for both of the ETMs so we can make a full comparison of the current confirguration we have. We excited in each LOCK stage at a frequency of 0.33 Hz (with no filters turned on), and took a transfer function of the oplevs in pith and yaw relative to the green arm locking control signal. The results are as follows:
ETMY:
Pitch (Mag urad/umeter, Phase deg) | Y (Mag urad/umeter, Phase deg) | |
M0 | (0.0336, 39.9) | (0.004, 164.2) |
L1 | (0.272, -125.5) | (0.0277, 30) |
L3 | (2.6, -179.6) | (0.017,63.1) |
ETMX:
Pitch (Mag urad/umeter, Phase deg) | Y (Mag urad/umeter, Phase deg) | |
M0 | (0.20(1), -19(1)) | -- |
L1 | (0.5(1), 36(2)) | -- |
L3 | (2.9(1), 0(1)) | -- |
Evan found no coherence between his excitations and the yaw oplev which is why those results are empty.
Conclusion: The ESD stage has the worst response for both ETMs in pitch, followed by L1. In Yaw, ETMY L3 and L1 are equally bad. ETMX seems to be worse than ETMY. This is a bit surprising since we only see green transmission drifts in the y-arm when we lock DIFF. One important thing to note is that we do not actuate on L3 at this frequency in ALS DIFF since the crossover between L1 and L3 is around 2 Hz as noted in alog.
For reference the templates can be found under: /ligo/home/sheila.dwyer/ALS/HIFOXY/ETMY_L2AngleNov72014.xml and /ligo/home/evan.hall/Public/2014/11/ETMX_L2P.xml
Kyle, Bubba Found the leak to be coming from the pump shaft -> Closed valve at pump inlet -> Using absorbent mats to soak up liquid (~1 gallon spill) -> Added water into chilled water system -> Will order pump replacement The output from this booster pump supplies the Kobelco compressor, AHU3 hallway taps (no longer used) and the LVEA (no longer used) -> This chilled water branch "dead heads" (no flow) when the Kobelco compressor shuts down as the solenoid water valves close when the compressor is deenergized.
Expect substantial 1-3Hz seismic noise from Hanford hauling on both day shift and swing shift today (11/7). No Saturday activity is expected. The daily numbers of truck trips in the current era are higher than S5 levels, especially on swing shift, and are comparable to or higher than S6 levels. I drove to PNNL on Stevens in the middle of the day yesterday and saw nearly a dozen dumpster-carrying trucks traveling either north or south in a 10-minute span.
3IFO: inventorying OpLev; transferring SUSs from temp containers to permanent ones; working on viewports; SEI assy started; Mitchell packing masses; bifurcation preperation for LVEA started (to go into effect on Monday) see LIGO Document M1400342-v2
Commissioning: Will start @ 10:00 AM today
SEI: Nothing to report. Business as usual.
CDS: work @ EX to end at 10:00
SUS: couldn't hear what was said.
Diff Power seemed to be creeping towards the high 9% range. The REFSIGNAL voltage was set to -2.04 from -2.02. The result is 7.5% diff power.
Alexa, Evan
In preparation for locking DRMI with reduced power on REFLAIR_B, there is now a BS1-1064-90-2037-45P (90 % reflector at 45° for p-pol) ready to go on ISCT1. It is not yet in the beam path. There is also a dump ready to go.
There is maybe some confusion about the polarization state on the table. LHO#7032 says it's s-pol, but (1) at least a few of the 45° optics on the table are marked 45P, and (2) we initially tried to install BS1-1064-90-1025-45S (90 % reflector at 45° for s-pol), and we got 10 mW / 27 mW = 37 % transmission rather than the expected 10 % transmission. We should go out there with a PBS and see what's going on.
After installing the s-pol optic, we did some resteering of the yaw on the REFL-M3 mirror (see D1201103) in order to recenter the beam on the PD. When we removed the s-pol optic, we steered it back.
Betsy, Travis
Attached are the results from the first round of testing of this QUAD. What we see:
1) Very poor coherence at low f which I had a hard time improving
2) Some cross coupling between T and R on the main chain
3) Some minor cross coupling between some P and L on the main chain
4) The 3rd L mode or 2nd P mode is split and the 2 are cross coupled on the reaction chain
Attached are the plots of this QUAD compared to the data from the other 3 3IFO QUAD units.
Note, damped TFs and spectra to follow.
2449.25 Hz line (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=14855) is visible in BS oplev PIT, but not in BS oplev YAW nor ITMX, ITMX, PR3, SR3, even when no LSC feedback goes to BS.
That line in OL signal becomes bigger when the IFO is in-lock even with the butterfly band stop, but the OL damp also seems to excite the motion at an amplitude that is comparable to the LSC signal through the band stop (attached right bottom).
Based on these observations, I moved the butterfly notch (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=14898) from M3 ISCINF to M2 coils so that everything is stopped (see the filter settings in the attached).
K. Venkateswara
Yesterday I had added a pre-amp to the temperature sensors on the Gnd_T240 and BRS. The gain increase should have been roughly 50 * 24/9 ~ 130. The attached trend seems to indicate this is correct as compared to the one in 14872.
The attached pdf shows the new ASD from last night. The noise seems to have only improved by ~2ish for the BRS and not much at all for the T240. A side by side comparison might differentiate sensor noise/pick-up/real temperature background though I suspect it is not real.
The ADC noise going to 0.5 mHz is described in Jeff's SEI alog: 614.
For the earlier case (gain of 0.2 Volts/Kelvin), the ADC noise was a factor of ~2 below the measured noise floor at 10 mHz. In the present case, the ADC noise should be a factor of ~200 smaller than measured noise at 10 mHz.
VM server hosting alog, awiki, SVN, and CDS-Bugzilla hung overnight. I captured the crash dump screen and logs this morning then restarted host and guests. Services restored around 9AM PT.
no restarts reported
K. Venkateswara
Sensor correction on X and Z at ETMX was left on last night. I turned it off this morning at 8 am as Filiberto and I were planning to work in the EX VEA. I modified the temperature sensors to take in +/-12V from the PEM power supply (thanks to Filiberto) and added a differential amplifier (with gain~50) to it.
In the afternoon, wind speeds were routinely hitting 50 mph. I've attached a pdf showing the ASD of the ground motion, BRS output and Stage 1 motion and some interesting coherences. Sensor correction was off in X and Z during this period.
Sensor correction has now been turned on and I will add plots of the result later.
J. Warner, K. Venkateswara
The X sensor correction we tried on ETMX was producing too much longitudinal motion in the X-arm at low frequencies. Sheila had to turn it off at 1:30 AM local time, this morning.
I've attached an ASD plot showing the GND_T240, GND_Super-sensor and Stage 1_T240 all along X. While SC improves Stage 1 motion above ~50 mHz, it injects a huge amount of low frequency motion. I think there are two main reasons for this:
1. While the super-sensor is likely measuring real displacement above 50-60 mHz, it is mostly measuring noise below ~30 mHz (see Brian's detailed SEI alog 602) . I'm still investigating the reason for this but it may not be an easy fix.
2. The SC filter is too aggressive (for the noise we have). Shifting it up to a higher corner frequency/faster roll-off may reduce low-frequency motion while hopefully keeping some of the benefit at 0.1-0.5 Hz.
Attempts are being made at better modelling. I apologize for rushing things without clearer understanding but hopefully, modelling and understanding will converge soon :)
This entry summarized the status of REFLAIR_B signals. REFLAIR_B is the broadband photodetector, D1002969, mounted in reflection of the interferometer which is responsible for measuring the 3f signals at 27 MHz and 136 MHz in DRMI. The 3f signals are relatively weak, especially at low modulation depth. This has lead to a low signal-to-noise ratio at sub milliampere photocurrent and distortion problems above a milliampere. The problem of saturation by out-of-band intermodulation products was recognized early on a LLO and fixed with a diplexer amplifier, D1300989.
The situation at LHO and LLO are essentially the same.
Koji has recently measured the second order intermodulation products of the RF chain in the broadband photodetector and found the second order intercept point to be rather low, around 30 dBm. This is maybe not too surprising considering these RF amplifiers are single transistors or Darlington configurations operated around a fixed DC working point. No specification of the IP2 is given in the datasheet.
This leads to the conclusion that 3f signals at LHO are predominantly due to intermodulation distortion in the RF amplifier chain and not due to the optical signals. This obviously works just fine for the DRMI, but probably doesn't give the required immunity to the carrier mode for full interferometer locking.
Possible solutions:
Increasing the modulation depth and reducing the photocurrent at the same time will not improve the situation, since the distortion depends on the absolute signal strengths of the individual RF lines. However, removing the first amplifier stage should give us enough room to further increase the modulation depth and improve the signal-to-noise ratio. Using a high modulation depth during locking may require an adjustable EOM driver, such as the, D0900760.
Summary
- Follow up measurement for the alog above was done.
- It was confirmed that the first preamp (MAR-6SM) is creating the domnant intermodulation and we will be able to improve it
by removing this first preamp as suggested (by costing some noise increase).
- It may become overkill if we are going to apply notch filters as being tested at LLO. Therefore it is also planned to test other amplifiers
that are similarly low noise to MAR-6SM, and are located between MAR-6SM and GALI-6 in terms of the intermodulation performance.
2nd-order & 3rd-order intercept points (IP2/IP3)
To quantitatively confirm Daniel's expectation above, I took measurements of the amplifier IP2/IP3.
IP2 and IP3 for an amplifier are defined by from the amount of harmonic distortions as
P2 [dBm] = P1 [dBm] x 2 - IP2 [dBm]
P3 [dBm] = P1 [dBm] x 3 - IP3 [dBm]
Here, P1 is the power of the linear output, and P2/P3 are the power of the 2nd/3rd harmonics.
When P1 reaches IPn, Pn becomes equal to P1. i.e. the output starts to be dominated by the n-th order.
Of course, we usually can't drive the amplifier at that level, this is purely a mathematical way to quantify nonlinearlity of the amplifier.
Basically the power of the bilinear intermodulation can also be estimated with IP2 in the same way as above.
Just replace P1 with the total power of two signals into the formula for P2. There may be some factors like 3dB, but just forget about it for now.
TEST1: Nominal configuration (MAR-6SM + GALI-6)
In order to measure IP2/IP3 of the nominal amplifier configuration of the BBPD, the input power was swept from -60dBm to -20dBm.
The input frequencies of 9MHz and 45MHz was used in order to check the frequency dependence. In fact, there was no significant
frequency dependence as we'll see in the result. Therefore only the input frequency of 45MHz was used in the other measurements.
Attachment 1 shows the relationship between the amplifier input power and the output power at the fundamental, 2nd harmonic,
and 3rd harmonic frequencies. The lines were manually applied to illustrate IP2/IP3. From the line for the linear power (red), the gain of
the amp chain was determined to be 32dB. In this configuration, IP2 and IP3 were 35dBm and 31.5dBm, respectively.
Practically, we want to know how much intermodulation (IMD) we produce when the amplifier is connected to the IFO.
I gazed Evan's measurement (14807) again and determied the combined power for 9MHz+36MHz, and 45MHz+91MHz to be
-0.5dBm (-32.5dBm at the input) and -11.9dBm (-43.9dBm at the input), respectively. These are indicated as the vertical black lines in the figure.
We expect to have -0.5*2-35 = -35.5dBm of IMD for 27MHz, and -11.9x2-35 = -58.8dBm of IMD for 135MHz. That is not too far from what we see
from Evan's meausrment. (Sanity check)
TEST2: The 1st preamp only (MAR-6SM)
Attachment 2 shows the same measurement only with the first preamp (MAR-6SM)
Roughtly to say, IP2 of MAR-6SM is reduced by a factor of 14.5dB, which is close to the gain of the second amp (13dB).
This means that the IMD performance of the chain is limited by this amp. Minicircuits show IP3 only in the spec sheet.
The measured value (18.5dBm) is close to the spec (18.1dBm). (I'm not insane)
TEST3: The 2nd preamp only (GALI-6)
Attachment 3 shows the same measurement only with the second preamp (GALI-6)
This amplifier has much better IP2/IP3 than the 1st one. Again the measured IP3 (38dBm) is close to the spec (35.5dBm)
This measuerement indicates that we'll have the IMD of -70dB and <-80dB relative to the source of the IMD when the first amp is removed.
Drawback & some other possibilities
As Daniel pointed out, the second preamp has worse Noise Figure than the first one. So we expect to have worse noise level in terms of the shotnoise intercept photocurrent.
Also Matt is testing on-board notch filters at LLO. If we consider to apply some notching, this GALI-6 could become overkill.
I ordered some other amplifiers like GALI-39, GALI-52 (Daniel's pick), and MAR_8A. They are similarly low noise to MAR-6SM, compatible packages
to the PCB, and located between MAR-6SM and GALI-6. Once they arrive, I'll carry out the same tests.
Alexa, Nic, Evan, Dan, Sheila
Tonight we went back to trying to get ALS DIFF working well. We aren't sure what the problem is.
First Nic checked that the gain of the 2 ESDs were the same. The old X gain in L3 LOCK L was 0.2, the new one is 1.12 The old Y gain was 0.7 the new one is 0.384. We also checked the crossover by using the green control signal.
Attached are some plots, based on the suspension model that Jeff used to design the DIFF loop, and filters downloaded from foton. This is using none of the boosts for DIFF, which is how we have been running the last few nights without being able to use the ESD. The two plots on the left are the cross overs for our loop, and the open loop. We can use these plots to compare what we have now to what Jeff originally designed, (G1400146-v2) We also copied the livingston filters, see alog (12590), their crossover is the third plot while their open loop (we added a scaling factor to get the right ugf) is the last plot on the right.
Btoh ESDs are driving, but we seem unable to use them in the loop.
K. Venkateswara
I had installed temperature sensors on BRS and GND_T240 yesterday as described in 14825. The first plot shows the trend over a day along with the PEM_VEA temperature sensor. The count to Kelvin conversion was expected to be 1.56e-3 K/count. This seems roughly consistent with the temperature of the T240. The BRS temperature sensor shows a lower magintude and a phase offset due to it's extra thermal shielding and larger mass.
The attached pdf shows the ASD of the temperature sensors and the coherence between them and their respective instruments. The temperature sensors are mostly limited by ADC noise. An op-amp based pre-amp of 50-100 gain would be useful. BRS_RY_Out shows a little bit of coherence near few mHz while T240 X, Y and Z show no coherence with the temperature sensor.
You should add the ADC noise to this plot
I've added a comment about it in 14909. I'm not sure how to display the ADC noise in DTT. In any case, it shouldn't be limited by ADC noise any more.
I wanted to try turning on HEPI sensor correction at EY this morning, but I've run into an issue. When I turn on just the new Mitt_sc filter no numbers come out of the outputs. However, when I turn on the filters associated with the fir sensor correction, numbers come out. Even when I just turn on the path, with no filters engaged, numbers come out of the output block. Something about engaging the sensor correction filter completely cuts the signal. I will try copying the filter to a different fm when I get a chance.
I screwed this up. Dave found that I had installed a filter with a gain of something like 10^-11, then fixed the gain, but never loaded the code into the front end. So foton showed a reasonable filter, but the front end was running a filter with a gain of zero. This is fixed now.