Thanks Jonathan, The golden question, perhaps not easy to answer, is knowing the threshold of operating environment for the power supplies and the chassis in current form.
Many users, I suspect, may not be aware or have considered operating environment or even cycle times. If they are aware of approaching boundaries where additional steps maybe prudent to ensure long term use of the equipment they can be prepared to adapt to the modifications you have suggested. Still, doing mods one or few at a time, as opposed the large quantity deployed, may have its own challenges. Finally, many thanks for offering to share your pains. Best regards, Mo Digicom 510-639-7003 _____ From: Jonathan Weintroub [mailto:[email protected]] Sent: Thursday, July 07, 2016 1:33 PM To: casper Cc: Mo Ohady; Alec Rust Subject: Re: [casper] ROACH2 Power supply issue An addendum to Alec's note on a power supply mounting idea: At SAO we adopted the FSP250-60FAG power supply for our latest build of ROACH2s currently in process at Digicom. We also had trouble with the iStar (sometimes labelled XEAL), and have had better experience with the FSP> We came up with an installation scheme which suits us-though certainly won't suit everyone. It involves fairly substantial mechanical modifications to the ROACH2 enclosure, and resulting machining expense. I thought I would share briefly what this looks like with pros and cons. If CASPERites want to use this scheme, we are as always happy to share drawings and experiences. The attached photos detail that we have machined out an aperture in the rear panel of the ROACH2 behind the power supply location, removed the power entry and EMI filter and line power pigtail, and moved the power supply back flush with the rear panel. Threaded holes in the power supply matched with through holes in the rear panel are used to screw it in place. We also machine away part of the top flange to which the enclosure lid screws. Pros: Eliminating the power entry and pigtail means fewer connections on the line power path, it's just simpler overall. The FSP mounts securely, the way it was intended to be used. The power supply vents directly to the outside of the enclosure rather than contributing heat to the inside. Cons: We are not as concerned as the enclosure designers (MeerKAT) with EMI compatibility-the mechanical mods almost certainly make the enclosure less RFI tight than it was spec'ed. Removing the EMI filter doesn't help in this regard either. Nor do the screw holes in the lid which no longer have mates in the flange. The original design of the enclosure used the part of the rear panel which we removed to mount up to 8 SMA RF connectors to bring analog signals into the enclosure. This modification is not compatible with this routing. There is additional cost per unit in the mechanical mods, likely greater for small production numbers. Jonathan On Jul 6, 2016, at 11:27 AM, Alec Rust <[email protected]> wrote: Hi all, SKA-SA have suffered a number of ROACH2 power supply failures. After investigation we determined that the power supplies (ISTAR TC-20FX2) overheat and then fail. This is due to obstructed airflow and fan failure. The front air intake of the ISTAR is almost completely blocked by the wire loom, and the fan is of the sleeve bearing type. In all the failed power supplies the fans had seized. One workaround is to set the ROACH2 enclosure fans to maximum, this will force some air through the power supply. Our failure rate dropped significantly after setting the fans to maximum, but we still encountered failures. This can be done by commenting out the following line in /etc/tcpborphserver3.init: #?job process r2fanctrl exec://r2fanctrl We have identified another power supply by FSP (FSP250-60FAG). We have not had any failures using this supply and more than 50 boards have been running for 3 months. The supplies do not have the correct mounting holes, so fitting them is an issue. We have used thin double sided tape, but this will not be sufficient if the LRU is to be transported. If you have any questions do not hesitate to contact me. Regards Alec Rust SKA-SA No virus found in this message. Checked by AVG - www.avg.com Version: 2016.0.7687 / Virus Database: 4613/12574 - Release Date: 07/07/16

