Summary:

The power supply and fan issues with the i/o chasses could be solved with a new power supply producing smaller magnetic fields (being planned by CDS), routing cables away from power supplies, and a separate supply for the fans.  The strongest magnetic field coupling sites to the OSEM read back channels were the cable connectors to the chasses that the signal passes through and not boards such as the a to d board. However, since the cables pass close to the power supply, the dominant coupling for the power supply fields was to the cables instead of the connectors at their ends. The separate issue of the fan frequencies showing up in channels can be solved by using an external power supply: the remaining coupling (from magnetic fields) when using a separate supply was measured here to be down by at least ten. 

Power supplies in the I/O chasses:

In  a previous log I reported that magnetic fields from the I/O switching power supply were strong enough to couple into channels at an unacceptable level (here).  Since then, I have investigated the magnetic field coupling sites to see if there were any coupling sites that could be mitigated.

I looked for magnetic field coupling sites to one of the UIM OSEM readback channels for ITMY. I used a centimeter scale coil and slowly scanned circuit boards etc., while monitoring the coupling on a screen. The strongest coupling that I found was to the cable connectors to the UIM coil driver, the anti-aliasing module and the I/O chasis. The coupling to the connectors was stronger than the coupling to the a to d board, or any other location inside the coil driver or I/O chassis or along the cables. Note that this is only a test of coupling to the read back channel: I have not fully tested coupling of rack magnetic fields to the coil actuators.

Notwithstanding that the strongest coupling site was to the cable connectors, the dominant coupling site for the fields from the I/O chassis power supply was to the cables themselves because they pass so close to the power supply. Figure 1 demonstrates cable coupling, showing that when the cable into the UIM coil driver was about 10 cm from a field generator set on the I/O chassis power supply, the injected comb showed up strongly on the channel, but much less so when the cables were moved about 40 cm away from the field source, with no other changes.

Fans in the I/O chasses:

I noted in a recent ilog (here) that i/o chasis fan frequencies showed up in the channels passing through the chasis. When the fans were run on a separate power supply, the coherence was lower, but because of time constraints and dtt crashes, I was not able to run long coherence measurements to make sure that the coupling was much lower. Figure 2 shows the results of a recent 79,000-average coherence measurement, and the fan peaks appear to be showing with a low level of coherence. The plot suggests that the coupling of the fans to the i/o chassis channels is down by a factor of about 10 when the fans are run on a separate power supply. 

For this reason, I think that running the fans on a separate power supply will be sufficient. So, to solve the issues I have raised recently, the i/o chasis power supplies should be replaced and the fans should run on independent supplies.  There is a chance that power supply ripple from the fans would not be a problem with the new power supplies. I have not tested this.