0 meter: big UPS of CB
600 meter: small UPS, connected by cable to IPS network of CB
1200 and 1800 meter: big UPS of 1500m intermediate building
2400 meter: small UPS, connected by cable to IPS network of NE building
3000 meter: big UPS of NE building
As a first test, we connected some voltage sniffers and a portable magnetic probe to the Onosokki. Results:
Fig 1: comparison UPS sniffer (CH2, magenta) and the IPS sniffer (CH4, blue). The 14, 36, 64,... Hz bumps are clearly seen on the IPS signal, but not on the UPS.
Fig 2: magnetometer placed on top of UPS machine (CH1, top plot), IPS sniffer (CH4, middle plot) and their coherence (bottom plot). Apart from the 50 Hz line, there is clear coherence for the 14 and 150+-14 Hz bumps, but not for the 50+-14 Hz bumps.
Fig 3: same signals as in fig 2, but with the magnetometer placed between the scroll and turbo pumps. No clear coherence except for the 50 Hz line.
Fig 4: same signals as in fig 2, but with the magnetometer placed on top of the vacuum electronics rack. No clear coherence except for the 50 Hz line.
Fig 5: same signals as in fig 2, but with the magnetometer hanging in front of power cabinet on tunnel wall. Broad coherence for all the bumps (not surprising, both sniffer and magnetometer see the same IPS signal).
We can conclude that the bumps are present locally at 2400N, but only on the IPS side, not on the UPS side.
Measuring with the portable magnetic probe and the Onosokki, we could indeed see the bumps near the power cables that exit the tunnel (no plots available). A few meter away, these bumps cannot be seen by the probe (which is not as sensitive as the big magnetometers).
Since the big magnetometers in the end-buildings have some problems with saturation, we brought one of the old ones from the central building, which are 'desensitized' by a factor 5. It was connected in the place of Em_MABDNE02, without adjusting the calibration value.
Events (all local time on 08/03/2012):
12:20 connected desensitized magnetometer from CB in place of MABDNE02
12:25 huddle test with MABDNE01 (east-west direction, close to west wall)
12:34 magnetometer placed east-west in middle of hall in front of entry doors
12:38 magnetometer placed east-west close to east wall, in mirror position of MABDNE01 with respect to tower
12:44 back to west side of the tower, in line with MABDNE01, but closer to tower. We left it like this.
Fig 1 shows the huddle test, showing the factor of 5.
Fig 2 compares the spectra for the huddle test (red) middle of hall (blue), east wall (green) and close to tower on the west side (black).
Fig 3 compares the spectra during the night with the 3 original magnetometers. The bumps at 14, 36, 64, ... Hz are well visible along the east-west and up-down direction, but are less clear along the north-south direction. This is consistent with a magnetic field generated by a cable in the north-south direction.
The evidence is not conclusive, but the fields seem a bit stronger on the west-side. This is consistent with a field generated by the IPS cable along the west-wall. The bumps seem to be the strongest close to the tower, which might be caused by distortion due the large amount of metal of the tower. This does not say anything about what is happening to the field inside the tower, which is what counts.
We tested different on/off configurations of the vacuum equipment (estimated power ~1500W) and the small UPS unit ("UPS-tube") at 2400N in the North arm tunnel. In addition, we used one test load (hairdryer, nominal power 1800W) which we plugged on different power lines: IPS or UPS-tube at 2400N, IPS or big building UPS ("UPS-building") at 3000N. For each configuration we compared the magnetic noise in the North End hall measured by the reference Virgo magnetic probe Em_MABDNE01.
Magnetic noise spectrograms are shown in Figures 1 (March 8th, morning), 2 (March 8th, aftrenoon) and 3 (March 13th).
List of events (UTC):
March 8th
09:20 2400N scroll pump off
10:10 2400N vacuum-rack off
11:51 2400N hairdryer on UPS-tube (till 11:54)
14:55 2400N UPS-tube off
15:03 2400N UPS-tube set in "by-pass mode" (by Massimo) and vac-rack switched back on
15:16 2400N lights on in the tunnel (till 15:20)
March 13
15:15 2400N vacuum-rack off
15:17 2400N ethernet hub board off
15:19 2400N hairdryer powered on UPS-tube in by-pass mode (till 15:21)
15:22 2400N UPS-tube switched off
15:25 2400N hairdryer on IPS (till 15:27, but tunnel lights were on)
15:30 2400N hairdryer on IPS (till 15:32, tunnel lights off)
15:37 3000N hairdryer on IPS (till 15:39)
15:40 3000N hairdryer on UPS-building NE (till 15:43)
Figure 4 shows the NE hall magnetic noise associated to the UPS-tube at 2400N (mod. Riello, 6.5kVA). The 14Hz structures are present when the UPS-tube is on with no loads (purple), they are amplified and/or shifted a bit in frequency when a load is present (black) and they definitively disappear when the the UPS-tube is switched off (green). It looks like some noise is also produced when the UPS-tube is set in by-pass mode (red).
CONCLUSION #1: 14Hz magnetic noise bumps in the NE hall are associated to the small UPS-tube unit atW 2400N (confirming hypothesis in 30956). We hypothesize similarly small UPS at 2400W, 600N and 600W generate 14Hz bumps in WE hall and CB hall respectively.
Figure 5: compares the NE hall magnetic noise associated to the (same) hairdryer plugged on different power circuits. (*Note*: the bumpy spectral structures are not corrupted data or windowing artifacts, but real glitches from the dryer. Figure 6 shows the effect was reproduced in the laboratory, 2-sec glitches are sensed by a magnetic probe next to the dryer or the dryer power cord, and as well by a voltage probe plugged at the dryer 220V outlet).
Hairdryer tests (Figure 5) showed that the induced magnetic noise was at the same high-level when we plugged the dryer on IPS line at both 2400N (blue curve) and 3000N locations (red). Instead, the induced magnetic noise reduced by at least a factor of 10 when we plugged the same dryer at 3000N on one power outlet served by the NE building big UPS (black).
CONCLUSION #2: in 30957 we found magnetic noise is radiated in the NE hall by the power cables inside the cable-tray running along the hall west wall. We can now more precisely attribute the magnetic noise intensity to these power cables geometry. Here it is: after inspection (and confir by M.D'Andrea) we now know this cable-tray carrys both one 380V-3phase IPS power line from the ENEL-to-1.5kV tranformer feeding the 2400N location, and one 380V-3ph UPS power line from the big UPS-building generator feeding the 3000N location. The "IPS power line" consists of four few-cm-thick separated cables (3 phases + 1 neutral) while the "UPS power line" consists of one single 1cm-thick cable in which the 3+1 wires are tightly packed together. A rough calculation indicates that radiated magnetic fields scale proportionally with the phase-neutral wire separation distance, the cm distance of the IPS wires versus the mm distance of UPS wires might well explain the factor 10 difference in radiated noise.
*Note*: the hairdryer test result is indeed confirmed by examining the magntic noise induced by 3000N and 2400N (equal) pumping stations: the 3000N station "uses" the UPS cable inside the west wall cable tray and its radiated noise is (at least) a factor ten less than that from the 2400N (equal) station that uses instead the "IPS cables" in the tray.
We hypothesize that excess magnetic noise in the West and Central hall is similarly radiated by analogous few-cm-distant phase-neutral cable pairs, although a careful inspection has yet to be carried out. A noise mitigation solution seems somehow feasible but is yet to be worked out. Future installations shall be carefully examined.
indeed, this is the complete author list. Thanks to all.