Search criteria
Section: X1
Task: DTS
J. Freed,
Spur Problem
As discussed in part 1, the Double Mixer has good phase noise around the carrier frequency to about a 100Hz out where SPI operates, however, it also has signals every 4096Hz away from the carrier. I had assumed that it was an issue relating to phase and amplitude mismatch on the phase delayer, but after adding attinuators to correct the missmatch, it helped, excepecially the 80Mhz + 4096Hz signal saw a ~15dB improvement. DM_Att.pdf. However, the double mixer still has large signals and the largest did not change with the addition of the attinuators.
I realize now that the problem lies with the mixers themselves as by the nature of a mixer they create spurs on the output.
DM_Mixer.pdf Shows the output of a single mixer. Along with the expected 80MHz - 4096 Hz and 80MHz + 4096 Hz there is extra peaks which i believe is caused by the Spurs of the mixer. The largest spur is at 80Mhz + 12,288Hz. This spur is actually in phase with the main signal and is thus construtivly interfered at the power combiner.
I do not know how to get rid of this spur as it is so close to the main carrier signal. For fun, I used a combination of Marki LC filter Design Tool and LTspice to simulate adding a filter to the end of the system to reduce the 80Mhz + 12,288Hz spur, but even an 8th order elliptical filter would only reduce the spur by ~0.04dB due to how close it is to the main signal. For fun I went to 20th order elliptic and it only reduced it by ~0.1dB. In addition, with nonstandard parts for a 20th order elliptic it got reduced by ~4dB. I am unsure of any other method to reduce the spur.
Double Mixer is a Single Sideband Mixer
I realize now that the basic design of the double mixer actually exists and is fairly common, it is called a Single Sideband Mixer (SSB mixer). Thanks to Brian, who gave me one of these mixers to test, I have something to compair the results
DMvsIRM.pdf Shows a plot of the double mixer vs a SSB mixer. The SSB mixer is tuned to 80MHz+4096Hz as apposed to the Double Mixers 80MHz -4096Hz but comparisons can still be made. For starters, the double mixers 80Mhz+4096Hz sideband equivalent is much better in the double mixer vs the SSB. However this comes at the cost of the 80Mhz+12,288Hz sideband equivalent which is much better in the SSB. In summary, just based on this single graph, the double mixer is more likely to be better for SPI for 2 reasons. One, we have already found out that SPI would be senstive to the 80MHz+4096Hz sideband equivalent. And two, if there is a way to attinuate these spurs by filtering, the spur further away from the carrier could be more easily attinuated.
High order modes filtering
Ive already discussed that adding fliters does not help the area around 80MHz. However, it does help with higher order modes
DM_Filter.pdf Shows a plot of adding a MC80-10-4AA bandpass filter from Lark engenering that has a bandwidth of 10MHz and a center of 80MHz at different positions in the Double Mixer. The best place to put one seems to be just before the amplifier, not entirly sure if adding one is nessisary to SPI but it should be good for reduceing a possible noise source.
High-Order Mode Filtering
The MC80-10-4AA that I listed actually filters by reflections, which is not helping to remove those signals from going back into the mixer. If a filter is going to be added, something like the ZXLF-K151+ would work much better.
I now realize a filter should be unnecessary as the AOMs are expected to have high insertion loss at frequencies not around 80Mhz, AOM-Tuneability.png, basically acting as a band-pass filter anyways.
J. Freed
Contiuning the work from 84198 this alog is about characterizing the other components of the Double Mixer. D2400315
D2400296 PCB Board transimpedance amp
According to oscilloscope, the signals coming from PCB board are from J2 being cos and J3 being sin. The Vpp are 1.199V on J2 and 1.195V on J3. As before the phase measures are hard to get a read on as the phase measuremnt on the osciloscope says 90deg when I put J3 in the channel 1 position and 91deg when I put J2 in the channel 1 position. Instead I put a 4096 Hz signal generator on Channel 2 then put the of the sin and cos signals on channel 1.
D2400296 PCB Board transimpedance amp With a 1V from DAC |
Relative Phase (deg)(+- 0.5) | Amplitude (dBm) (+- 0.1) | |
Cos(J2) | 152.7 | -30.6 | |
Sin(J3) | -117.7 | -30.6 | |
Difference | 89.6 +- 0.7 | 0 |
ZLW-1BR Mixer
The mixer data was collected at a range of values incase LO input power needed to be changed, however I would not recomened going below 4dBm LO as the isolation of all the ports falls gets worse the lower the power. The indivitual signals were colected on the Agilent 4396B with a BW of 10 Hz (best one the spectrum analizer allowed without giving an overload notification). While the Total power was given on the E4418A power meter. SRS SG382 was used for the LO power.
ZLW-1BR RF out power with a 1V IF input | 3dBm LO power | 4dBm LO power | -4dBm LO power | -7dBm LO power |
At 80 MHz +- 4096Hz (dBm)(+- 0.1dBm) | -4.0 | -3.4 | -10.8 | -13.8 |
At 80 MHz (dBm)(+- 0.1dBm) | -39.5 | -38.0 | -46.2 | -49.2 |
Total (dBm)(+- 0.01dBm) | -0.89 | 0.25 | -7.52 | -10.51 |
ZMSC-2-1BR Combiner
The combiner values were colected by using the Agilent 4396B to apply a sweep of values around 80 MHz. All values were the same in about a 10kHz span that was measured
ZMSC-2-1BR Combiner with -1 dBm on the port | Relative Phase (deg)(+- 0.05) | Power Drop (dB) (+ 0.01) |
Port 1 | -18.71 | -3.12 |
Port 2 | -18.65 | -3.11 |
Difference | 0.06 | 0.01 |
ZMDC-10-1+ Couplier
The couplier was characterized by applying a -1dBm from a SRS SG382 signal generator signal at 80MHz on the input port and reading off from the E4418A power meter. The other port not being measured was terminated. We do not care about relative phase so it was not collected
ZMDC-10-1+ Couplier Powers at 80MHz | Power with -1dBm on Input (dBm)(+- 0.01) |
Out | -1.61 |
Couple | -12.50 |
J. Freed
The reason we picked the ZFL-500HLN amp for SPI's Double Mixer was that it was an amp that met requirements from minicircuts
-frequency range contains 80MHz (our operating frequency)
-Operates at voltage supplied by Low Noise Power(15V, 17V, 24V etc.)
-Has a P1dB well above our output power of 10dBm (to remove non-linearity)
-No heat sinks (to fit in a 1U chassis.) (This requirement actually doesnt matter, the ones discussed here are still the best even compaired to amps with heat sinks)
-sma connections
-Low noise
-Low gain (we only need ~8dB of amplification)
The ZFL-500HLN was chosen because it fit all those requirements, with the cavieat of having to attuniate before amplification (adding noise). Though if I am correct and the Low noise power board 5V Vcc port is capable of supporting a 63mA load, another option would be the ZX60-P105LN+ instead due to the lower gain and lower noise. This would remove the need for an attinuator before amplification but its low input return loss means alot of power is reflected back towards our mixers (Also causeing noise). So it becomes a pick your poison of adding noise by attinuating before amplification vs by reflections back into our mixers. In practice, I believe both would operate about the same. Will add a comment if something changes
Double-MixerAmpComparison.png shows a quick comparison of the 2 amps
Acording to this term definitions manual I found on minicircuts:
"Directivity (active) is defined as the difference between isolation and forward gain in dB. It is an indication of the isolation of the source from the load, or how much the load impedance affects the input impedance and the source impedance affects the output impedance. The higher the active directivity (in dB), the better the isolation."
So directivity is just a measure of Isolation - Gain. Which isolation is just a measure of how much power is sent through the amp to the input port with power applied on the output. This would really only affect the amp when there is a large impedance mismatch on the output (load). If there is little to none, directivity doesnt really matter. But if it did come up, the ZFL-500HLN would do much better. The greater concern I believe for the ZX60-P105LN+ is the large input return loss, which is the power of the reflections off of the ports back the way it came. This would cause a standing wave on our input. To reduce it, we would have to add an attinuator on the input anyway. Probably less than the 10 dB attinuator which we currently have, but still something to consider.
P.S. VSWR and Return Loss(RL) are the same thing with different units, the conversion between them is VSWRvsRL.png. A VSWR closer to 1 is better, while a higher return loss is better. Converting the ZFL-500HLN aproximate 1.04:1 VSWR to Return loss gives 34.15dB. This means the reflections power are about 20dB greater on the ZX60-P105LN+. Really not sure if it is worth it excpecilly since mixers are sensitive to these reflections.
J. Freed
Last week, this week, and the next week, the goal is to finalize the double mixer design, drawings and chassis. M. Pirello. has begun the drawings while I finalize the design, and we will work on the layout as well as the front plate.
Currently 83439 shows that phase noise is good except for harmonics every 4096Hz away from carrier. While any frequency more than 100Hz around the carrier should not affect SPI some of them do have an effect (namely the 8192Hz harmonic), 81593 also shows an ocsiloscope reading of the output which shows how messy the signal looks. This as such this week the focus is on reducing those harmonics.
The Q2 transistor on the Low Noise Power Board failed last week, was replaced, then it popped and replaced again. The problem was diagnosed to be the power pins on the ZFL-500HLNB+ amp. The pins stick out and touched the metal casing of another minicircuit part causing a short on the +15V supply from the low noise power board. A temporary solution of electrical tape was use to isolate the pins while a more perminate isolation for the pins will be there when it is finalized
AGBW.jpg Shows that the Agilent 4396B in spectrum analyzer mode gives less information with a smaller resolution bandwidth. The higher resolution bandwidth shows more harmonics with greater power on said harmonics. I do not know why this is, which is why I will be looking at phase noise when trying to improve harmonics.
The Agilent 4396B in network analyzer mode was used to take transfer functions of the double mixer phase delayer, IMG_1472.jpg shows the set up. The source power was 13dBm, the central frequency was 80MHz. The Span and Resolution Bandwidth were both 40kHz but changeing these values did not change much as the amplitude and phase was relativly flat around 80 MHz. A 10dB attinuator was added to the output to remove the overload message on port B that appered around 7dBm input power when I was incresing power to 13dBm.This is ok as I could calibrate the machine by replacing the phase delayer with a addaptor that let the signal go straight through. This calibration became my (0dBm 0deg) reference. I took measuments of one port of the phase delayer at a time while terminating the other.
Gain (dB) +- 0.01 | Phase (deg) +- 0.1 | |
Port 1 | -3.31 | -84.6 |
Port 2 | -2.93 | 3.1 |
Difference | 0.38 | -87.7 |
This is slightly off from what was hoped for with (0dBm -90deg) difference. SPI has some 0.5dB attinuators as well as 1dB attinuators. Due to the attinuators not being their listed value at 80 MHz, Two 0.5dB attinuators on port 2 and one 1dB attinuator on port 1 seems to nearly correct the power differnece.
Continuing from here is characterizing the mixers, the summer, and RF couplers to gain inisight on reducing the harmonics.
J. Freed
Today I collected 2 phase noises, first phase noise is of the 80 MHz OCXO after it has gone through an RF distributer. The reference was the SRS SG382 at 80MHz
The second is of the Double Mixer again but with an extra focus on the 8192Hz sideband.The reference was agian the SRS SG382 at 80MHz-4096Hz
These tests was done with the direct plug in method without using a PLL. This requires both the device under test and the reference device to be at quadrature. Normally a PLL is required to keep them there, however since both devices are connected to the LIGO timing system in some way, a PLL is unnecessary to keep them locked in quadrature. To put them in quadrature I just lined them up on an oscilloscope best I could then added 90 deg by adjusting the phase on the SRS. I bypassed the internal amp of the BluePhase1000 for all tests.
DM_vs_RF.pdf Is a plot of the two phase noises, as seen, the low frequency noise (<2Hz) has these humps. I now attribute these to the RF distributer causeing a shape change in the pure sin wave signal.
RF_Distributer_Signal.jpg Shows how the RF distributer has a not exact sin wave (Yellow) compared to the SRS Pure tone (Blue)
DM8kSideband.pdf Shows a close up on the 8k sideband that will also couple into our SPI output
The measured curve looks very similar to the SRS SG382 spec at 100MHz. There is no reason to believe the SG382 is better than the OCXO.
J. Freed
Before the hoidays as shown in 81658, the Double Mixer had large phase noises at 4096Hz that were thought to be caused by phase offest difference in the Double Mixer. Thanks to Marc Pirello for the suggestion and help to manually adjust the phase difference by extending the wire lengths inside the double mixer.
DM_PN2.pdf Shows the lower 4096Hz peak in the path adjust(green) as compared to the no path adjust (blue) and both compared to an independant SRS (orange). The harmonics are still large so DAC output had a ~5kHz low pass applied but with negledgable improvement. Investigation was ongoing when I went on holiday.
Today I plugged in the double mixer into a network analyzer to see the power spectrum density of the signal
DM_NA_WIDE.pdf Shows that although the DM has better noise floor, the harmonics of the double mixer are quite a bit worse in the Power spectral desnity than shown in phase noise DM_PN2. The best result (shown) was with NO added length. Unsure why this is.
DM_NA_NARROW.pdf Besides that the signal has a fairly pure tone, though this is limmited by the resolution.
Investigations ongoing
J. Freed,
Update on Double Mixer Progress 81593, Proceded through step 2a of the double mixer test plan T2400327. Initial results are sugesting a possible noise improvemnt compaired to the other options in the area of interest for SPI.
DM_PN1.pdf Shows the phase noise test run in step 2a of the Double Mixer Test plan. While not a true 1-to-1 comparison of the phase noise performance of the double mixer compered to other options (step 2b is for that), it shows that adding the double mixer into this system (except for the peaks) improves phase noise performance by a factor of 2-5 from 100 Hz to 20 kHz. Of note, there is a large peak are centered around the 4096 Hz that is of interest to SPI. As there was only a cursory attept to properly phase match the signals in the internals of the double mixer for this initial test, the 4096 Hz sideband was not properly removed.
A possible cause of this phase miss match is our phase delayer inside the double mixer (ZMSCQ-2-90B) causes a phase delay of about 89.82 degrees at 80MHz and not the 90 degrees we are expecting.
Possible fixes
The very low frequency (<0.05Hz) contains DC noise caused by the external phase mismatch of the double mixer and the reference source for the phase noise measurments. It is not an indication of double mixer drift and there has not been yet investigations in drift.
J. Freed,
On Wednesday Nov 20th, the Double Mixer topology for SPI D2400315-x0 was taken, and it works how we intended as a way to gerenrate a 80MHz - 4096Hz signal it to but with alot of noise.
DMPrototype.png Is a diagram of the Double Mixer Design
DM_First_Test.png Is a diagram of the test done on the Double Mixer, this inital test was just to check that the Double Mixer produces the expected output signal frequency, as such, the amplifier was ignored.
DM_FT_Result.png Shows the output on the scope. Blue is the double mixer output. White is a reference of SRS SG382 (sync with 80 Mhz OCXO via 10 MHz Timing) with a frequency of 79,995,904 Hz and power of -0.91 dBm. The SRS signal, timed with the OCXO, and the double mixer correctly locks with eachother. This gives credance that this design will work for the goal of the double mixer; a 79,995,904 Hz signal frequency locked to a 80MHz OCXO. However there is a lot of noise in the double mixer signal. An ongoing investigation is currently underway to find and reduce this noise.
ZMSCQ_Wave_Shape.png Shows a possible source of the noise, the Double Mixer requires phase delayer for the Sin path, we used a MiniCircuts ZMSCQ-2-90B for this. I tested the ZMSCQ-2-90B by puggining the input of the signal and the signal from the 2 ports into osciloscope. I split off the input signal with a Minicircuts ZMSC-2-1BR+ (Summer/Splitter) and I had to normalize input power pickoff in postprocessing by a factor of 0.68 due to the power difference. The results show that the output of the ZMSCQ-2-90B has a sligtly different shape than what was plugged in. With the delay port (orange) having a slight "hump" on the rise. Will retake with a faster osciloscope, as the signal is very noisy.
The investigation is continuing with the next step of checking the double mixer output signal with a network anaylizer to check the frequencies around 79,995,904 Hz. A Phase noise measurement would also work with the caviot that the phase adjust to destructivly interfere the double mixer signal with the reference signal would have to be adjusted manually.
J. Freed
I followed Step 1 of the double mixer test plan in T2400327 under "1. Characterize frequency references."
PNSPIDACTestStand.pdf Shows the results in dBc/Hz. And PNSPIDACTestStandrad.pdf shows the result in rad/sqrt(Hz).
Besides the 80MHz standard reference there were 2 devices under test, the IFR 2023A and the SRS SG382. These two devices went under 3 tests involving different combinations of
1. Timing to a 10MHz frequency standard produced by the 80MHz Standard. (Time Standard)
2. PLL locking to a different 80MHz frequency ref through the tune in port. (Lock)
During this process it was assumed that the 80MHz standard and the added 80MHz ref would have a simmilar noise profile as they come from a simmialar OCXO.
A 4th test was added to the Time Standard/No Lock test which used the SR785s internal high pass filter (-3dB at 0.16Hz) to remove some of the DC components from the test. This is because the results of the initial test were inconclusive as the noise floor of the SR785 was too high. The noise floor was too high because there was a DC signal that caused the input range of the SR785 to be about 20dB. The high pass filter, removed the DC signal, lowering the noise floor at the cost of signals below about 0.2Hz not being accurate. We believe this is caused by the fact that while the PLL locking can be controled (tries to lock signals to destructivly interfere so to lower output power for Phase noise tests), there was no control by the Time Standard on phase differences between oscilators. As Time Standard/ No lock, showed the best results above 0.06Hz and is closest to the SPI set up design, this fact may be important when doing futher tests.
Between IFR and SRS, the SRS showed better phase noise performance below 7kHz. While IFR showed better performance above 7kHz. As such, SRS shows more promise in SPIs phase noise range of interest 0-4096Hz.
After a talk with Jeff what is really important to show is the TIme Standard/Lock which is the 80 MHz OCXO, IFR and SRS which we believe is most represntative of phase noise
These are, one at a time, locked to another 80 MHz REF OCXO. In this it was assumed that the 80 MHz OCXO and the 80 MHz REF OCXO would produce the same phase noise, as such contribute equally to the output phase noise. As such the output was manually halved to represent the only the 80 MHz OCXO. As stated before these measurments also had a Time Standard (sometimes called time sync) where the 80 MHz OCXO is synced to the site's 1PPS. And the IFR and SRS have a 10MHz signal time standard signal that was produced by the 80 MHz OCXO divided by 8 and sent into the back of the devices.
PNSPIDACTestStandrad(sync_lockOnly).pdf Shows only the results of the Time Standard/Lock, which we believe to be the most represntative of the phasenoise of these devices.
Since it was assumed that the two OCXOs would have the same phase noise, this also makes our noise limit equal the OCXO. As such, it can be seen from this graph that we are at the noise limit in measuring the phase noise of the IFR and SRS between 0.1Hz-10Hz.
J. Freed, M. Pirello, J. Kissel
SPI Double Mixer D2400315 prototype is now built and ready to be tested T2400327, in the SPI DAQ test stand D2400283
Oli Patane, Jeff Kissel
Continuing through our plan(T2300376) for the new O5 A+ suspensions (HRTS, HATS, and BBSS), we have finished, compiled, and installed the Simulink models for HRTS and BBSS onto the Triples and BSC teststands respectively. The models that we installed can be found in /opt/rtcds/userapps/release/sus/x1/models/x1sus{bs,lo1}.mdl.
Also, we are close to done creating the medm suspension overview screens for HATS and HRTS (/opt/rtcds/userapps/release/sus/common/medm/hxts/SUS_CUST_{HRTS,HATS}_OVERVIEW.adl). They still need some editing, but here are the current state of the medm screens for the HRTS and the HATS.
Rolf, Keith, Carlos, Dave:
During the CDS Face2Face meeting at LHO June 6-9, we worked on the LHO DTS:
I needed to work on some code on the test stand and found that x1ldasgw1 had restarted last Friday ~23:48 localtime. So I had to remount the frame directories for x1nds1 and x1fw1.
Dave and I had been discussing https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=36294. One solution we thought about was just disabling the optimization that gave us bad data. The thinking was that when it was put in we used 1 minute second trend frames. It would be faster to open one 10 minute second trend frame than ten 1 minute trend frames. I wanted to characterize the impact of the shared table of contents optimization.
NDS1 variant | Timing | Notes |
---|---|---|
Current production code | 0.723s | Fast, but broken around fw restarts |
No TOC optimization | 2.077s | Slow, but always works. |
TOC optimization w/ knowledge of fw restarts | 1.595s | Medium speed, but always correct. |
My thoughts on this are:
1. Getting correct data from nds1 is important, restarting the daq should not be a reason to give bad data (even if it is only for a few minutes).
2. We should preserve the table of contents optimization, it does have some impact.
All timings are from my laptop connecting to the LHO DTS x1nds1 server via an ssh tunnel. It should be noted there were several restarts of the nds/fw/dc on the DTS during this time window, so the penalty of not being able to share the table of contents is high in these short samples. I ran the queries below multiple times to make sure the disk cache was hot and picked a representative time for each case.
The measurements:
Quick query against x1nds1 with the current production code (no changes)
$ time nds_query -d 400 -n localhost -p 8088 -s 1179600000 -e 1179603000 X1:DAQ-DC0_DATA_RATE.mean,s-trend
Data for 600 seconds starting at GPS: 1179600000
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179600600
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179601200
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179601800
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179602400
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
real 0m0.723s
user 0m0.008s
sys 0m0.000s
Quick query against x1nds1 with the frame table of contents optimization disabled.
$ time nds_query -d 400 -n localhost -p 8088 -s 1179600000 -e 1179603000 X1:DAQ-DC0_DATA_RATE.mean,s-trend
Data for 600 seconds starting at GPS: 1179600000
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179600600
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179601200
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179601800
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179602400
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
real 0m2.077s
user 0m0.000s
sys 0m0.008s
Quick query against x1nds1 with using the table of contents optimization with some additional knowledge of fw restart times. This is were we need to be, always correct, but able to use some frame access optimizations. This can be improved by taking into account which restarts are also channel list changes, and which are just restarts.
$ time nds_query -d 400 -n localhost -p 8088 -s 1179600000 -e 1179603000 X1:DAQ-DC0_DATA_RATE.mean,s-trend
Data for 600 seconds starting at GPS: 1179600000
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179600600
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179601200
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179601800
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
Data for 600 seconds starting at GPS: 1179602400
Channel type nWords units
X1:DAQ-DC0_DATA_RATE.mean real_8 600
real 0m1.595s
user 0m0.004s
sys 0m0.004s
Jonathan, Dave, Jim:
At 03:04 Sunday 5th February the DTS SAM-FS server (x1ldasgw1) rebooted itself unexpectedly. The daqd process on x1fw1 was not running at the time, so no data was lost. We restarted SAM-FS and NFS services and rebooted x1fw1, all is now recovered.
The mouting of /cds-h2-frames took longer than usual (several minutes as opposed to a few second).