2002 March 07 From J.D.Monnier Subject: Ground Problems at IOTA Dear all. This message may no longer be relevant, considering that Mike and Rafael eliminated most of the detector noise. However, I thought I would pass along what I learned in case it every might be useful to anyone. I had a conversation with my colleague Sunil Golwala (Caltech) who has a lot of experience with low-noise measurements and has encountered (and solved) numerous problems involving bad grounding. After talking with Sunil, I discussed things with Ettore and learned a few more things about the situation at IOTA which I didn't know before. I want to summarize my (limited) state of knowledge here and to receive feedback on whether some of the 'facts' are indeed true or not. My current thinking: [could definitely be wrong] 1. The so-called "Clean" and "Dirty" power at the IOTA site are confusing and misleading misnomers. The "Clean" power is simply power going to the Uninterruptable Power Supply (UPS) while the "dirty" power is not on the UPS. We should have been calling this the "UPS" power and the "non-UPS" power. 2. Dirty and Clean power are traditionally meant to separate the ground returns for sensitive (low-signal) applications [like IR cameras or any detection system] from other electronics which deal with large amplified signals. What this means is that ALL the computers, motors, anorad cage, etc. should be on DIRTY power since an unstable ground or noise on the AC will not change cause any problems. The only thing that should be on the CLEAN power at IOTA are the IR camera and MAYBE the ccds. At IOTA, almost everything is on "UPS" power because we don't want the computers to go dead if the power to the mountain fluctuates. This is logical, but means that everything important shares the same return lines which means the IR camera is extremely vulnerable to ground problems due to fluctuating current demands by the telescope motors, the delay line motor currents, the VXWorks, etc. 3. Ideally, we should have two UPSupplies, one for the true 'Clean' and one for true 'Dirty' power, each well grounded to the same earth-point. Perhaps we can actually do this, since we have low power demands for true 'CLEAN' electronics; maybe we can buy an additional relatively cheap UPS.. However, the wiring and new grounding might be difficult and expensive [maybe not -- we could put all the current 'dirty' wiring on the old UPS and all the current 'clean' wiring on the new one..?] 4. Sunil thought that 10s of Khz - 10s of Mhz spikes are more consistent with ground problems than radiatived RF-pickup. Stuff at the kHz range tends not to feed into your clean electronics because the electronics will have voltage regulators that will reject low frequency stuff (unless it's huge). As a rule of thumb, simple analog electronics (stuff built with op-amps) tend to crap out in the 100s of kHz and up -- the op-amps stop acting like infinite gain differential amplifiers. He pointed out that copper braids are poor conductors of high frequencies (due to skin effect) and that one might need to ground things better. He recalled a similar problem which was solved by grounding a rack to earth-ground using a 1-2 ft wide copper sheet -- the extra surface area eliminated the high-frequency ground problems by offering lower resistance. They typically used thicknesses like 1/32" to 1/16" -- thick enough to be sturdy, but thin enough that you can bend and manipulate it. When you bolt it to something, use lots and lots of screws (more than look reasonable) because the sheet will only work as well as the weakest interface. Finally, Sunil recommended we take a look at the power supply lines in your electronics, meaning the post-regulator supplies that actually power the circuits on the boards. If we see spikes here, that's a good bet that it's power supply feedthrough. Another test is to power the IR camera off some car batteries temporarily and see if this cleans things up. 5. Other advice on RF shielding from Sunil: "We try to make our instruments totally hermetic, with sheets of metal rather than just aluminum foil. But aluminum (or, better, copper) foil can be used temporarily while you figure out what you want to do. You just have to make sure it really connects to the surfaces it touches -- use lots of conductive tape. Another thing to remember is that aluminum oxidizes immediately on contact to air, so aluminum-aluminum joints tend not to be that great. One can gold-plate things, or there's a cheaper process called alodining (not anodizing!) that will prevent aluminum surfaces from oxidizing. Standard materials that don't oxidize easily are copper and brass. There are also electrically conductive RF sealing gaskets that one can use to seal the doors of boxes that are opened frequently." ------- Possible Actions to eliminate extra noise on camera: 1. Temporarily connect all the anorad/vxworks/tlescope motors/etc ("Dirty" equipment) to "non-UPS" power and make sure the camera stuff is on UPS power. Since these (hopefully) have different returns we could see if the camera problems still occur. While this is not a permanent solution since we are vulnerable to AC power problems, this might tell us something. 2. We could try grounding the anorad/vxworks rack better using copper sheet. Also we could try grounding the camera itself to ground better using a copper sheet so that the faraday cage can exchange current with ground to shield out RF. 3. Look into a new UPS supply.. This might be a reasonable thing to do no matter what and might eliminate other problems like how the local internet is vulnerable to noise when certain motors run, etc.. 4. I think Mike Brewer's plan to find and eliminated the SOURCE of high-frequency current spikes in the anorad cage is good and should still be pursued. 5. In case the problem is caused by radiated RF pickup, we should try some shielding experiments. I personally don't know how much shielding is necessary, but I think we should try shield the cabling from the camera dewar to the A-D card cage (and the cage itself) better as a test at least. Also from Sunil: "You can get ground braid that cables can be strung through. Another thing to think about is how the cable shields are connected. Typically, a cable will have an internal shield, and one can connect it to the backshells at each end (which should be metal and make direct contact to chassis at each end). It's important to have a completely hermetic system. For example, if you have a cable running from your instrument to your computers, and its shield is only connected on one end, it won't work well as a shield because the resistance along the shield is nonnegligible. Also, check the pinouts of your instrument cables. Are all the analog signals in a separate cable, or are analog and noise (digital, power) signals running in the same cable? Are the analog lines twisted pairs and do they have a shield between them and any nasty signals?"