This page: http://monitor.mikes.fi/ionclock/
Machines and IPs:
Strontium/linux .45.99 (strontium.ad.vtt.fi) (192.168.10.1) - since 2021-12-03 IP = .45.15
win: .45.23 (ESPTE1M012-ioni.ad.vtt.fi)
kavi: .45.190 (kavi.ad.vtt.fi)
comb: .45.15 (192.168.10.18)
N210 SDR: (192.168.10.20)
News and notable results
2022-08-26..29: Menlo comb DDS measurements
2022-08-12: 2nd AOM-chirp measurement
2022-06-18: New OPA818-based PDH photodetector with lower (shot noise limited?) noise
2022-05-21: first search for Autler-Townes splitting
2022-05-02: first AOM chirp measurement (measurements 2022-04-28/29)
2022-03-04 (59642) to 2022-04-03 (59672) ROCIT-campaign
Changed PSU from Thorlabs LDS12B to side-of-cavity banana-plugs for the 357 MHz fiber-noise detector
1 kHz and 500 Hz signals (old PSU) are probably artefacts from the (helper) ARTIQ 178.5 MHz + 500 Hz DDS-signal interfering
Possibly a small change in the 50Hz and 100Hz signals when servo is off? (but a new peak at 22 Hz or so??)
- freerun1k, AOM-signal from ARTIQ, with 500 Hz offset giving 1 kHz beat-note
- freerun2, repeat
- vco, AOM-signal from VCO, servo not locked (drifts quite a lot)
- locked, servo locked
- servooff, servo electronics power off, noise floor
Servo/VCO locked: level is far below noise-floor when servo is off?
500Hz peak remains - why?
VCO unlocked shows 1 kHz broad spectrum with side-bands at 1.5-2 kHz
noise-floor with servo off shows small 50Hz and 100Hz peaks!?
noise-floor large peak at 1 kHz is the same as closed-loop 500 Hz peak??
PDH detector signal with spectrum analyzer
- yellow: SA floor
- green: detector dark
- blue: detector bright
- red: locked to cavity
measurements through 10dB coupler
data: CSV file with same name as image
Measure PDH-signals and fiber-noise signals with Sound-Blaster x-fi
- no (large) 50 Hz component in instrument noise-floor or actual signals
- PDH error-signal similar when laser unlocked and laser off (not much evidence for shot-noise-limit here?)
- clear 34 Hz peak from A/C
- audio-digitizer not very useful <10 Hz ?
- reference 357 MHz as usual
- VCO replaced with ARTIQ DDS at 178.5 MHz + 546 Hz
- after double-pass this gives 1092 Hz beat-note on error-signal
- spectrum plotted with -1092 Hz frequency offset
Add grounding-wire between FALC and DLCpro, this seems to reduce the 10ms spikes a lot (completely gone?)
Measure with FFT: (all traces 1601 pts, frequency 0 Hz to 120 Hz)
ch1: FALC monitor when laser locked
ch2: sesmometer Y, X, Z
FALC signals also with FALC-off, and with FALC-on but no input signal connected
Errorsignal (FALC-input) via BNC-T to FFT-analyzer
- laser locked
- laser unlocked
Measurement of 50 Hz peak with FFT-analyzer seems hard/impossible, since peak remains when FALC off and/or input-signal disconnected
34 Hz signal seems 'real', from KSK A/C-machine
Not much evidence for 100Hz modulation
Measure FALC-Monitor signal with Keysight black scope, Mini-Circ LPF filter (11MHz?) cuts high frequencies
Try FALC on/off states, and FALC-ULI -> DCLPro Fine-in-1 cable connected/disconnected
Connecting the Fine-in-1 cable gives spikes with 10ms period in FALC-monitor. Not good!??
Measurements with no photodetector-(error)signal connected to input of FALC.
FALC power on.
FALC power off.
For comparison, scope-scale increased 500-fold to 500 mV/DIV, FALC-monitor when laser is locked:
Toptica win-program settings:
PDH freq 25 MHz, Amplitude -18.2 dBm
Input level -10 dBm, Lock level 0V, Phase 199.3deg
Piezo ARC: Fine-in-1, -0.5 V/V
Try also KSK105 effect? N=5 averaging mode, so takes some time for spectra to change.
15:54 KSK105 on as normal
15:58 KSK105 off
16:00 no 34 Hz peak in Y-signal
16:01 reduced 34 Hz peak in Z-signal also. 30 Hz peak remains
40-41 and 43-44 Hz signals remain - some motor?
16:14 KSK105 back on
17:12 stop data collection
Clear reduction of peak at 34 Hz (Z stronger, Y weaker) as well as 57.5 Hz (Z,Y equal)
Spectra when turning off devices (oscilloscope off already): Siglent Spectrum-analyzer, Siglent PSU, black PSU, Thorlabs PSU, Oscilloscope on top of fft-analyzer
>5 dB reduction in 50Hz Y-peak - otherwise no major changes
Oscilloscope on top of cavity-box causes 75Hz peak of height 20-30dB
14:13 start on new file: 20220927.txt, 10s interval for spectra, no averaging
75 Hz peak possibly from Rigol oscilloscope on top of cavity-box? (oscilloscope now off)
14:15 try oscilloscope on
clear peak at 77Hz initially, then 76 Hz, then ca 75 Hz - fan in oscilloscope?
14:18 oscilloscope off
14:18:20 spectra: 75 Hz peak disappears
14:21 spectrum-analyzer (Siglent) off, also on top of box
14:23 EOM RF-amplifier + orange fan off (on floor at input-end of cavity, RF-amplifiers needs fan-cooling?)
14:26 black PSU ("Aalto") off, powering cavity through-transmission power servo
14:29 Siglent PSU off, powering fiber-noise servo, 674nm trap-breadborad-power servo
14:33 small rigol-scope on top of FFT-analyzer off
14:40 TK203 (kallion välitila, tuloilmakone) off (no VFD, so runs at 50 Hz?) - no major change in spectrum?
Atlas Copco compressor shows 88 Hz acoustic peak with spectroid
14:44 TK203 back on
14:47 Thorlabs-PSU LDS12B (on floor) off, powering fiber-noise photodetector(?)
Similar model PSU causes bad B-field near ion (was used for RF rectifier, at some point)
Y-signal 50Hz peak drops to <-90 dBm, slight improvement?
strong remaining Z-peak ca 43-44 Hz!? (compressor??) and 34 Hz (KSK105), and 57.5 Hz (??)
14:51 turn on N=5 averaging on FFT
Note no averaging during bottom half of image. Top half with N=5 averaging in FFT-analyzer.
Calibration waveform applied to FFT-input around 12:00
Trillium compact voltage output sensitivity is ca 750 V * (s/m)
measurement is velocity, need to differentiate to get acceleration?
Old IonClock project cavity acceleration sensitivities estimated:
0.15 e-10 s^2/m longitudinal
0.13 e-10 s^2/m transverse
0.24 e-10 s^2/m vertical
From ion-clock spectra, estimated 50/100Hz modulation
Beta_50Hz = 0.9
Beta_100Hz = 0.55
FM modulation depth (?) 0.9*50Hz = 45 Hz ? -> 1e-13 fractional
1e-13 / 0.24e-10 (s^2/m) vertical = 4e-3 (m/s^2) vertical acceleration amplitude a_ampl at 50Hz
a_ampl = vel_ampl*2*pi*f_mod
-> vel_ampl = a_ampl / [2*pi*f_mod] = 1.3e-5 (m/s)
Test of FFT-analyzer scale.
Signal from Keysight signal-generator (trap drive), set to 45 Hz.
Keysight: Oscilloscope: FFT-ch2-peak:
10 mVpp ca 20 mVpp -30 dBm
5 mVpp ca 11 mVpp -36 dBm
1 mVpp 3.4 mVpp(noisy) -50 dBm
dBm = 10*log10( [ Vampl/sqrt(2) ]**2 / Z / 1 mW )
Vampl = Vpp/2
Note Keysight Vpp is into 50 Ohms, but we use FFT-analyzer with 1 MOhm input impedance.
Vampl = sqrt( 10**[ dBm/10 ] * 1 mW * 50 Ohm ) * sqr(2)
Peaks in Z-spectra Vampl Vel_ampl=Vampl/ [750 (V*s/m)]
75 Hz -75.8 dBm 5.1e-5 V 6.8e-8 m/s
44 Hz -80.3 dBm 3.1e-5 V 4.07e-8 m/s
Peaks in Y-spectra Vampl vel_ampl
50, 75, and 100Hz all at ca -82 dBm 2.5e-5 V 3.35e-8 m/s
Seismometer spectra from night, on minus-k, doors closed
Effect of wedges?
- Z higher noise around 60 Hz, lower noise around 32-35 Hz
- Y slightly higher noise around 70 Hz, slightly lower around 30-35 Hz
Minus-K peak that was around 2.4 Hz (+ weak 2nd harmonic) last week, now around 10Hz (note also 2nd harmonic 22Hz) ??
from cables? or some details related to Minus-K floating?
remaining high peaks around 32Hz, 50Hz
16:22 seismometer power was of?, turn on.
16:33 doors closed
16:34 spectra with doors closed
Try KSK105 off? ca 16:40-16:42
ch1 autorange on?
16:47 spectra ok?
start new file: 20220926_2.txt
Note Minus-K trace in these plots probably wrong (Seismometer had no power?)
New file: fft_test_10s_20220926.txt
Seismometer on stone from start of file 09:45 or so.
10:19 turn off seismometer
10:21 FFT-analyzer electrical noise-floor with Seismometer off, but cables connected.
10:28 seismometer on concrete, FFT-analyzed Z/ch1 input range changed
10:29 first spectra on concrete, OK
10:33 ch1/Z input range back to sensitive setting (was overrange before)
10:35 ch1/Z to -14 dBm input range, ch2/Y is -4 dBm
13:56 seismometer off
move onto stone
14:00 ch1/Z shows overvoltage as transients settle?
14:06 ch1/Z range to -6 dBm, averaging off
highest Y-peak now ca 72 Hz, broad peak around 70 Hz
14:10 N=5 averaging on
14:12 averaged spectra ok? (now ch1/Z settled, 'half' light off at -6 dBm setting)
14:14 to -12 dBm rage ch1/Z ('half' light on)
14:15 averaged spectra ok
14:18 remove wedges?
14:46 most wedges removed (one of 8 wedges broke, some of it remains (middle, towards back of lab)
14:47 ch1/Z range now -18 dBm (half-light on steadily)
15:00 seismometer off
15:02 on minus-K, floating
ch2/Y to -14 dBm range
Effect of wedges seems not very clear/good? Need to repeat no-wedges measurements?
Seismometer on stone, from Saturday evening to Monday morning
Autocal-off on FFT-analyser seems to help keep labview data-collection running
file: fft_test_10s_20220924.txt (332 Mb)
fft file: fft_test_10s_20220924.txt
had to reset fft-analyzer, some settings probably changed...
0 - 120 Hz Frequency, 10s record length, 1 MOhm input impedance
FFT is average of 5
data collection interval now 20s
17:56 Seismometer on concrete
18:00 fix X-axis sot that frequencies start at 0 Hz
18:03 turn off seismometer
18:05 reset measurement - Seimsometer now on stone
18:07 first FFT collected, on stone
18:10 move seismometer to lab floor
18:11 reset FFT
18:12 first FFT collected, on floor
install wedges under stone!?!
18:16 Z-signal was overrange, chane input to avoid overrange
18:21 seismometer off
18:27 first FFT, on stone, with wedges
auto-cal off on FFT-analyzer (this causes timeout in labview-program?)
18:33 leave lab
8mm Hex-key fits between stone and concrete, 8.5mm metal-block does not fit
34 Hz peak identified as coming from KSK105 (set-point 35 Hz)
Datasheet: Trillium Compact
new fft file: fft_test_10s_20220922.txt
13:44 electrical noise-floor of FFT-analyzer, with nothing connected to BNC-inputs
no major peaks at 50Hz or 75Hz remain
13:47 50R termination on both input channels
13:48 initial transient after conencting 50R-terminators dies down. Z-spectrum slightly better comperade to nothing connected to ch1. Y-spectrum(ch2) similar.
13:50 back to seismometer signals.
13:53 open cavity door (large transients)
Y-signal recovers quickly back to similar spectrum as seen before
Z-signal takes long time to settle? FFT-analyzer overvoltage/oversignal?
14:03 set Z/ch1 to AC-coupled, now similar FFT-spectrum as before
no averaging on FFT-analuzer
14:09 seismometer on stone, next to ion-pump controller
sharp spikes in Y-signal?? 2Hz repetition rate, two spikes close together?
14:14 spkes disappear??
time-series shows large oscillations, especially Y-signal, at ca 50 Hz
spectrum shows highest peak at 43-44 Hz
seems like stone has a resonance around 40-50Hz? (oscilloscope image from hitting stone with hex-driver)
14:26 FFT signals connected again. Y-spectrum high (>-60 dB) peak at 43 Hz (stone resonance??)
ca 44 Hz and 2nd-harmonic 88 Hz also highe peaks in Z-spectrum
14:34 seismomenter down to concrete pillar
Z-signal now has 2 Hz spikes (internal mechanism of seismometer, after it is moved?)
Y-spectrum settles faster, seems ok 14:36, only 50Hz spike remains
14:42 signals look strange, so reboot seismometer (DC supply off, on)
14:45 averaging-mode, N=5, on FFT-analyzer
16:30 data from 14:45 -> ca 16:30 should be fairly quiet, from concrete pillar
ca 68Hz peak visible now also?
16:42 seismometer on lab floor, input-side of cavity
peaks 40, 45, 49 Hz
16:48 reboot seismometer (Y-signal was at +10V?)
16:49 spectra look settled
Z-peak at 24-25 Hz, also 34 Hz (from KSK105?)
17:00 back to concrete pillar
17:04 -> FFT data looks ok?
Note 68Hz extra signal appears in afternoon ca 15 local (similar to yesterday)
Seismometer on stone (next to ion-pump controller). Response when hitting stone at other end, with hex-key handle
Note response has frequency in 40-50 Hz range!?
11:25 -1floor PK304A off (this motor caused problems 2022/02, damping towards ceiling and horizontal tubes improved 2022/02)
spectroid on phone shows vibrations around 30-35 Hz from this motor? (VFD set-point is 35Hz)
ca11:32 PK304 back on
11:40 -1floor PK102 off (set-point probably 50 Hz)
ca11:47 PK102 back on
Spectroid shows strong 42Hz signal close to AtlasCopco compressor (-2floor)
from spectrum looks like one problematic motor set-point could be 44 Hz? (motor vibrates with some 'slip' from set frequency, maybe -1 Hz lower)
also one peak lower at 41-42 Hz?
LabView program file-writing a bit unstable, data collected until ca 19.
time-axis is not completely linear (program stops sometimes), but time-stamps should be accurate
17:20 -3floor KSK105 off
17:30 -3floor KSK105 on, 35 Hz
17:47 scroll pump on
18:02 turbo soft start
18:20 turbo full speed
image color scaling min=-130 dB, max=-75 dB
Average spectrum for all data, and spectrum with KSK105 on/off (ca 23 spectra, 10s/spectrum, averaged for each setting)
Seimsmometer installed at side of cavity
Minus-K Z-direction resonance frequency adjusted to maybe 0.8 Hz, should be floating now.
Seimsmometer Z-signal to ch1 and Y-signal to ch2 of FFT-analyser
FFT-analyser set to 0-120 Hz frequency and 5x-averaging (check?)
Traces read to file over GBIP every 10 seconds
Intermittent strong signal around 68Hz - source unknown?
From ca 17:20 local to 17:30 local, KSK105 in -3floor was off (normal setting is 35 Hz lately)
this is seen as lower noise around minute 80 in the plot.
SDR-tracker ch2 LO-mixing frequency
Set to 215 MHz
CH2 frequency in MU: 60517119992791
artiq CH2 actual frequency in Hz: 214999999.999999851 (this is probably wrong? check that FTW is even!)
In ZMQ-receiver code (takes gnuradio output and writes to InfluxDB) we have:
ch2_mix = 214999999.999996304512 # from even AD9912 tuning word: 60517119992791-1
Result of first counter-test with pi, lambda, omega counting.
Some issues with numerical offset in pi-counting still?
Test SDR-software with outputs for pi, lambda, omega counting.
Chronograf dashboard "SDR counter weights"
ca 17:20 measuring output with no FM
17:35 start FM-modulation loop
FM-depth 50 kHz
modulation frequencies linspace(0.01,3,55)
1000 s sleep at each modulation frequency.
Frequency-response test of SDR-tracker+counter, using 1 s gate-time
Measurement as described in Dawkins2007, which gives abs( W(f) ) from measured ADEVs with FM modulated input-signal
Note Vernotte2016 (17) is square of measured response.
Weight-function (for freqeuncy vs. time data) for Pi, Lambda, Omega counters, and corresponding frequency-response
Try to characterize beat-note SDR-tracker.
Signal from Keysight (trap-drive) generator
f0 = 14.024 MHz
fmod = adjustable 0.1 Hz to 3 Hz ?
FM-depth = 50 kHz (quite large?)
Amplitude 700 mVpp (seems OK for N-radio?)
fmod = linspace(0,3,55)
lowest possible modulation frequency seems to be 1 uHz
start loop over modulation-frequencies at ca 17:30 local
1000s per modulation frequency
signal connected to SDR-channel 1
modulation frequencies are linspace(0,3,55):
[0. 0.05555556 0.11111111 0.16666667 0.22222222 0.27777778
0.33333333 0.38888889 0.44444444 0.5 0.55555556 0.61111111
0.66666667 0.72222222 0.77777778 0.83333333 0.88888889 0.94444444
1. 1.05555556 1.11111111 1.16666667 1.22222222 1.27777778
1.33333333 1.38888889 1.44444444 1.5 1.55555556 1.61111111
1.66666667 1.72222222 1.77777778 1.83333333 1.88888889 1.94444444
2. 2.05555556 2.11111111 2.16666667 2.22222222 2.27777778
2.33333333 2.38888889 2.44444444 2.5 2.55555556 2.61111111
2.66666667 2.72222222 2.77777778 2.83333333 2.88888889 2.94444444
ca 17:45:30 change to 2nd modulation freq fmod 0.055 Hz
2022-09-06 ca 9:25local: turn off FM modulation
AOM-chirp weekend run
- when switching to a new attenuation, the RCDAT is a bit slow, so we end up with a bit of the old attenuation during the first repeat
- 0dB first repeat was with warm AOM, second repeat has gain adjustments, 3rd+4th should be ok.
- 2...18 dB seems ok
2022-04-28: first AOM chirp measurement. Limited SNR due to 60 MHz detector, ARTIQ attenuation level 18...28 dB
2022-08-12: programmable attenuator RCDAT installed, larger attenuation range. Better SNR with old PDH-detector (modified for higher bandwidth)
2022-09-05: optical setup shielded from aircurrents with cardboard + plastic bag shield
Back to AOM-chirp measurements.
ARTIQ DDS source at 76 MHz and 15 dB attenuation.
RCDAT adjustable attenuation from 0 dB up
B-SDR tuned to 151999576.99998077750205993652 Hz
400 Hz software-shift
Resulting beat-note at 23 Hz, recorded at 200 S/s
Friday 15 to Monday 07, ca 64 hours.
Each run is 30min on, 30min off.
Attenuations from 0 dB down to 28 dB
2 dB steps give 15 different attenuations [0:2:28]
-> 4 repeats per attenuation gives 60 hour measurement time
ca 14:56 local: start AOM-chirp run
15:57 lower SDR gain to 14dB (spectrum was strange looking before this) (start of 2nd repeat with 0 dB)
CEO DDS measrement at setting 35 MHz - 1 Hz
Predicted (nHz) Measured (nHz)
-223.517 -228.4 (32.0)
Updated DDS quantization figure:
CEO DDS measurement at 35 MHz + 1 Hz setting:
Predicted (nHz) Measured (nHz)
+223.517 +222.1 (32.4)
Results so far:
Leave same measurement running with DDS-setting of 35.000001 MHz over night
Expected fact-fset(35 MHz + 1 Hz) is +223.517 nHz
(For an offset of 35 MHz - 1 Hz, the expected fact-fset is -223.517 nHz. this could be measured also to validate DDS model)
Result from 16h CEO_DDS measurement with Microsemi 3120A:
35 MHz output mixed down with 42 MHz LO from ARTIQ -> meausurement around 7 MHz (low-pass filter before phase-meter)
DDS-model with CLK=14*10 MHz and 48 bits predicts zero frequency-error for 35 MHz output
f_LO = 42000000.000001586973667 (as output from artiq, check for numerical accuracy?)
f_measured = 7000000.000001602 Hz (stdev 7.762e-6 /sqrt(N=54900) = 3.313e-8 Hz)
Final DDS-error measured is: -13.7 (33.1) nHz - compatible with prediction of zero.
DDS quantization measurements so far:
Yet another attempt at measuring the 35 MHz f0/ceo DDS
Mix down with ARTIQ 42 MHz (as below)
IF amplified with mini-circ amplifier (9 dB attenuator to avoid compression - some 2nd harmonic still present)
11 MHz LPF on IF-signal to avoid 35 & 42 MHz at input of phase-meter
Meausrement at 7 MHz with Microsemi 3120A phase-meter (+2.5 dBm signal as reported by Microsemi)
NOTE: ceo-SYNCRO DDS has setting of "14-mult"! this may mean 14*10 MHz DDS clock!?
ca 16:25local start 16h run
NOTE2: mult=14 giving 14*10 MHz DDS clock seems to give exact output at 35 MHz!
CLK = 140 MHz
bits = 48
DDS_out = CLK * FTW/(2**bits)
gives exact FTW corresponding to 35 MHz, since 35/140 = 1/4
Try measuring DDS-frequencies with N-SDR.
ch1: rep-rate DDS directly at 19.999854 MHz
ch2: mixed-down ceo DDS at 35.000000 MHz
Mixing LO from ARTIQ, 42 MHz nominal
in MU: 11821949021848
actual frequency in Hz: 42000000.000001586973667
-> ch2 measurement nominal 42-35 MHz = 7 MHz
NOTE: set mixing-frequency to zero in zmq-receiver (writes to influxdb), needs to be reset to old value before clock-run
valid data from ca 11:25 local into database
long cables to SDR, may require stable lab-temperature to get sane results?
16:10 local stop measurement. N-SDR measurement has significant bias at ~1 uHz level - seems not suitable for measuring 0.1 uHz DDS offsets currently.
Try verifying f0/ceo DDS frequency using HP 53230A counter.
see gist read_counter.py on github?
Measurement has lots of noise and averages down very slowly - try N-SDR measurement instead?
20 h measurement of rep-rate DDS frequency over weekend.
DDS setting 19 999 584 Hz.
Frequency offset from nominal / nHz
Predicted 16h-run 20h-run
-171.363 -173.7 (5.9) -165.4 (3.0)
weighted mean -167.105 (2.67) nHz
around 16local: leave 60 hour run with campaign setting of 584 Hz. Should finish by Monday morning.
Summary of DDS quantization results
Actual DDS-frequencies are spaced 426 nHz apart. Figure shows quantization error for integer Hz settings around 19.999584 MHz
3120A phase-meter measurement of rep-rate DDS-output (back of comb-rack)
Reference is 10 MHz output from the menlo 'PLO'-unit, behind the screen
('584' is the setting used in the ROCIT campaign)
DDS-setting (Hz) Predicted offset (nHz) Measured offset (nHz) Menlo-screen-indication Predicted Comb-shift@1348nm
19 999 576 +18.626
19 999 577 +100.583
19 999 578 +182.539
19 999 579 -160.187
19 999 580 -78.231
19 999 581 +7.451
19 999 582 +89.407
19 999 583 +171.363
19 999 584 -171.363 -173.7 (5.9) "-1 Hz" -1.71e-16
19 999 585 -89.407 -83.0 (35.1) "-1 Hz" -8.94e-17
19 999 586 -7.451 -8.5 (25.6) "-1 Hz" -7.45e-18
19 999 587 +78.231 +73.2 (13.7) "0 Hz" +7.82e-17
19 999 588 +160.187
19 999 589 -186.265
19 999 590 -100.583
19 999 591 -18.626
clk = 12*10 MHz (10 MHz input, 12-fold clock-multiplier)
bits = 48 (48-bit tuning word)
f_dds = clk*FTW/2**bits
1. compute exact FTW based on desired setting: FTW_exact = f_set*(2**bits)/clk
2. round to integer binary numer: FTW_int = round(FTW_exact) (or perhaps floor() or ceil()?)
3. compute actual output f_act = clk*FTW_int/2**bits
4. compute DDS error: f_act - f_set
The spacing between actual DDS-frequencies is clk*1/2**bits = 426.326 nHz
if we always round the exact ftw to the nearest integer the maximum error is 0.5*426.326 nHz = 213.163 nHz
Comb-lock: 1 GHz (4th harmonic) rep-rate compared to maser-referenced 980 MHz + f_DDS
-> rep-rate error is DDS-error/4
for 1348nm we measure against comb-mode n_comb = 889558
f0 = -35 MHz (from similar DDS as rep-rate!)
f_IR = f0 + ncomb*frep - fbeat(ch1) (**)
Evaluate fractional shift from DDS-error (rep-rate):
f_IR = f0 + ncomb*(245 MHz + DDS/4) - fbeat(ch1)
DDS = f_set + f_err
this gives an f_IR error of:
f_IR_err = ncomb*f_err/4
y_IR_err = ( ncomb*f_err/4 ) / (0.5*SRS)
The prefactors together are 1e-9 [1/Hz], so:
y_IR_err = 1e-9 [1/Hz] * f_err [Hz] (compare to numbers above, -171 nHz DDS-shift gives -1.71e-16 fractional comb-shift)
For f0/ceo a similar DDS is used.
The model predicts a +141.561 nHz error for the 35 MHz DDS (set frequency of 35.000000 MHz) (so far we did not verify this with measurement)
This contributes with opposite sign, because Eqn (**) has -35 MHz
The shift at 1348nm is y_IR_err_ceo = -141.561 nHz / (0.5*SRS) = -6.37e-22 (negligible compared to other errors)
In conclusion it seems the ROCIT campaign used a DDS-setting where the comb-mode(s) used are shifted -1.71e-16 from nominal.
A correction of +171 (e-18) should be applied to data in post-processing.
From a 16 hour phase-meter measurement this number is verified as +174(6) e-18.
Lenovo laptop, USB3 to SDR. IP xx.xx.45.170
new artiq dds frequencies
ch1 MU: 70368718281966
ch2 MU: 70368718563441
Sanity-check of SDR measurement
Replace comb rep-rate input with ARTIQ-generated DDS signal.
ch1 nominal 249.999942 MHz
actual: 249999941.999998748302460 (MU: 70368727852115)
ch2 nominal 249.9999 MHz (no change from past runs)
actual: 249999899.999998807907104 (MU: 70368716030166)
SDR tune 249999509.99678924679756164551 (no change)
SDR file 2022-08-24_check-1.dat
run2 with new SDR tune
SDR file 2022-08-24_check-2.dat
run3 with bandpass filters around measured frequencies.
ch1 bandpass around 42 Hz (37-47 Hz, 5 Hz transition)
ch2 low-pass around 0 Hz (5 Hz cutoff, 5 Hz transition)
run4, frequencies closer together?
CH1 frequency in MU: 70368717437541
CH1 actual frequency in Hz: 249999904.999999225139618 (artiq does not round to even FTW!!)
CH2 frequency in MU: 70368716030166 (no change)
CH2 actual frequency in Hz: 249999899.999998807907104
SDR tune: 249999494.99874386191368103027
ch1 SDR-note at +10 Hz
ch2 SDR-note at +5 Hz
Reconfigure sampling and filters on SDR:
250 kS/s Hardware sample-rate
10 kS/s 1st-stage decimation (25x). Low-pass filter 4 kHz (80% of 5 kHz Nyquist-freq, transition 500 Hz, 10% of Nyquist)
1 kS/s 2nd-stage decimation (10x), Low-pass filter 500 Hz (transition 50 Hz)
200 S/s 3rd-stage decimation (5x), Low-pass filter 80 Hz (transition 10 Hz)
Result from overnight run, with DDS setting of ROCIT campaign
SDR-tuning is different, giving larger 800 nHz offset for ARTIQ-reference
Final result is similar to previous value ca +200 nHz
Change settings to ROCIT-campaign values for overnight run
DDS 19 999 584 Hz
SDR tune nominal: 249999510 Hz
SDR tune actual: 249999509.99678924679756164551 Hz
start of run 16:41 local
Comb DDS-setting 19999587 Hz.
Rep-rate measured from M-NIR branch with usual beat-note photodetector. Measures ca +220 nHz high from nominal
ARTIQ reference frequency measures ca +13 nHz high (note swings both below and above 0 due to temperature-changes)
Difference frequency ca +210 nHz high - corresponding to 8e-16 shift at SRS/2 (comb mode-number ncomb=889558)
Measurement from Monday afternoon to Tuesday. Image shows residual from linear fit to phase vs. time measurement.
Note significant slopes at start of measurement, from lab-temperature changes(?).
[Nemitz2021] notes on comb rep-rate measurement.
fr = 250 MHz comb rep-rate
fHM = 100 MHz maser reference
Maser multiplied by m=92 (Nexyn Corporation NXPLOS)
Comb harmonic n = 37 (DSC40S photodetector, Discovery Semiconductors)
n*fr = 9 250 MHz
m*fHM = 9 200 MHz
These mixed together, beat-note at 50 MHz to counter.
Counter accuracy requirement relaxed by factor n*fr/50MHz = 185
Dual channel rep-rate measurement using Artiq DDS as reference, with settings:
CH3 set frequency MHz: 249.9999
print:CH3 frequency in MU: 70368716030166
print:CH3 actual frequency in Hz: 249999899.999998807907104
print:CH3: delta f, 1.1920928955078125e-06
Comb DDS at 19 999 587 Hz
Comb rep-rate 250+(20e6-DDS)/4 = 249 999 896.75 Hz
measure +2.749 uHz offset from expected rep-rate? (ca 6000s dataset, after throwing away first 3000s of data)
2nd run: +2.879 uHz
Try comb rep-rate DDS setting of 19 999 587 Hz. This shows zero error on Menlo RRE-syncro??
17local start phase-meter run of length 24h
restart SDR-run, 2022-08-18-comb-reprate-200Sps-2.dat
SDR-tune 249999499.99809250235557556152 + software-shift +400 Hz
Monday 2022-08-22 stop SDR data collection.
analysis gives 8.9856e-8 Hz offset
Some offset seen between expected and measured comb rep-rate?
4th-harmonic of rep-rate is compared to phase-locked 980 MHz reference.
Resulting 20 MHz IF is phase-locked with programmable DDS.
This gives actual rep-rate as:
frep = 245 MHz + fDDS/4 = 250 MHz + (fDDS - 20 MHz)/4
Reprate-DDS Phase-meter SDR
??? check divider??
19 999 584 Hz +278(63) nHz +278.17 nHz
Test RF-lock of comb to maser.
Rep-rate set to clockrun-setting of: 250 MHz - (20MHz - 19 999 584 Hz)/4 = 250 MHz - 104.0 Hz
Rep-rate measured with usual beat-note photodetector (clock-laser is off) on the M-NIR comb-branch.
250 MHz beat-note measured with two instruments:
250 MHz signal to Hittite HMC394 5-bit programmable divider, set to 25-fold division.
Divider output amplified with mini-circ amplifier to get suitable level for phase-meter.
Measurement at 10MHz with Microsemi 3120A phase-meter.
Expected result is (250 MHz - 104 Hz)/25 = 10 MHz - 4.16 Hz
directly at 250MHz with B210 SDR
SDR-tuning at 249999499.99809250235557556152
software shift of +400 Hz
Samples recorded at 200 Sps
Verify electrical RF-power delivered to AOM during chirp tests.
Signal chain: ARTIQ 15 dB, AA PA, RCDAT adjustable attenuator -> AOM
RCDAT SA-measurement at 76 MHz
60 -31.5 dBm
20 +8.2 dBm
fit to this data gives RF at AOM: +27.96 dBm - RCDAT-attenuation
Set new SDR-tuning frequency: 151999029.99972695112228393555
10 dB RCDAT attenuation data collected during night
15 dB data during day.
15:45local For the last (?) repetition of 10, install plastic bag and cardboard shield to reduce airflow over beam-paths
run continues with 20 dB and 25 dB repeats during evening/night
16:25local start new SDR-file for the 20 dB run
15:50local start of 10dB attenuation aom-chirp run (note first repeat differs from others)
16:00local start chirp run, (1800s on, 1800s off) * 10 repeats * 6 attenuations [0, 5, 10, 15, 20, 25].
SDR datafile ca 6 Mb/hour * 60 hours = 360 Mb
Friday 4pm to Monday 9am is 65 hours. Time to do 6 different AOM attenuations, 10 hours each
Working attenation range is ca 0 dB to 25 dB, in 10 steps
ARTIQ attenuation set to 15 dB
mini-circ att, SA peak height (optical beat-note at 152MHz, via splitter to SA)
0 dB -13
24 -55 (SNR only ca 10 dB now)
2022 Spring season archive: index_2022spring
New AOM-chirp measurement setup, now with mini-circ programmable attenuator after the AOM PA.
Old PDH-detector was modified with 3k9 transimpedance to give bandwidth suitable for measuring double-pass beat note at 150 MHz
Setup is similar to 2022-04 (2022-04-28) chirp-measurement, with added adjustable mirror to align beam onto detector
PDH-detector is followed by mini-circ amplifier (ca +23 dB) and splitter to SDR and Siglent-SA.
SNR for beat-note with maximal AOM-drive is >65 dB - clear improvement from old setup with 60MHz detector
142 Hz sideband @ -51 dBc
1.04 kHz sideband @ -44 dBc (this could be mechanical resonance of some component - sideband seems to jump up when hitting on the trap-breadboard?)
Note 0.4 nW floor of power-meter restricts measurable range to ca 0-20 dB attenuation.
Measured power seems to agree with model, using optimal RF-power P0 = 1.3*Pmax_used
single-pass model is I/I0 = sin**2 [ (pi/2) * sqrt(P/P0) ]
small-angle approximation sin(x) ~ x for small x.
-> I/I0 ~ P single-pass, and I/I0 ~ P**2 double-pass
When this scaling holds, we expect 10 dB of reduction in optical power (OD1-filter) for each 5 dB of reduced RF power.
With RCDAT inserted after PA, 0dB setting, now ARTIQ attenuation ca 15 dB seems OK.
0 3.8 uW
6 605 uW
9 177 nW
14 19.8 nW
off 0.4 (artiq DDS off)
Test 674nm power vs. AOM RF drive power, old configuration (no adjustable attenuator)
674nm power-servo set-point to 2.0 V, lab lights dark, detector-area covered with holy-blanket
Thorlabs slim detector after 461-cube, before B-fiber
Detector dark: 0.2 nW
18 dB DDS att 4.03 uW
19 3.45 uW
29 105 nW
30 67 nW
DDS off 0.5 nW
18 4.00 uW
AA Optoelectronics whitepapre on diffraction efficiency:
I1/I0 = sin**2 [ (pi/2) * sqrt(P/P0) ]
this formula for single-pass, we use double pass.
Note RF PA may go into saturation at high power-levels!
Earlier measurement of the RCDAT programmable attenuator: RCDAT-3000-63W2
Measured over typical AOM frequency range of 60-90 MHz.
Lowest dataset is SA floor
Migration to ARTIQ-7 done.
Starting migration to ARTIQ-7
Summer break 2022
For 2022 Spring results see index_2022spring
For 2021 results see index_2021
For 2020 results see index_2020
For 2019 results see index_2019