[evlatests] December Engineering/Tests Meeting
Alan Erickson
aerickso at nrao.edu
Mon Dec 19 17:16:31 EST 2016
If we view the EVLA as a testing facility for RF switches, we've had 308
devices under test for (conservatively) 58k cycles with 24 failures,
from which we can calculate the observed MTBF as (308/24) * 58k = 750k
operations. Given likely errors in the number of wear-producing
operations, I consider this comfortably within the manufacturer's 1000k
MTBF -- quite so, as this data encompasses the infant-mortality stage.
Although we should ensure that our sparing and maintenance costs are
accurately forecast, I don't see this as an alarming failure rate.
Please holler if you think I'm off in left field.For those who would
like to see a short refresher on MTBF, Vicor has a digestible summary
(MTBF on p. 7):
(note: direct pdf
link)http://www.vicorpower.com/documents/quality/Rel_MTBF.pdf
On 12/19/2016 12:40 PM, Bryan Butler wrote:
>
> found a bug. conclusion still the same.
>
>
> time period SBs hours band switches no band switch
> ------------------------ ----- ----- ------------- --------------
> 2016 nov 1 - 2016 nov 30 266 560 1290 12605
> 2015 dec 1 - 2016 dec 1 3214 5912 13902 170043
> 2012 dec 1 - 2016 dec 1 12126 24250 58227 684849
>
>
>
> Bryan Butler wrote on 12/19/16 12:00 :
>>
>> some data on how many band switches we make.
>>
>> numbers are approximate because i'm not including test time (though we
>> don't often switch bands during testing), and i don't catch quite
>> *everything* that is science observing, but they are first order
>> probably about right. i did three scales of length of time as a sanity
>> check that the values scale about right, which they do.
>>
>> over the month of november 2016 we had ~570 hours of science observing
>> time in 274 SBs on the VLA. during that time there were ~1942 true band
>> switches, and ~12400 scans that didn't switch bands. i give "~" values
>> because i'm assuming that the first scan of an SB always switches bands,
>> which isn't necessarily true, but i figured was the most conservative,
>> and it will catch some of the missed test time band switches.
>>
>> over the year from december 1 2015 to december 1 2016, we had ~5912
>> hours of science observing time in 3214 SBs on the VLA. during that
>> time there were ~20800 true band switches, and ~163000 scans that didn't
>> switch bands.
>>
>> over the four years from december 1 2012 to december 1 2016, we had
>> ~24250 hours of science observing time in 12126 SBs on the VLA. during
>> that time there were ~86300 true band switches, and ~656780 scans that
>> didn't switch bands.
>>
>> steve tells me that hichem says that if there is no change in bands at a
>> scan boundary, the switch is not told to do anything. if that is really
>> the case, then these switches have not gone through anything even close
>> to 1000000 cycles. if, however, for some reason they actually are
>> cycling at such a scan boundary, then we're getting to about 3/4
>> lifetime.
>>
>> -bryan
>>
>>
>> Rick Perley wrote on 12/14/16 15:24 :
>>> All:
>>>
>>> We have three items (so far) for discussion. There's time for more
>>> -- let me know if you have something to show/tell/discuss.
>>>
>>> 1) Steve: RF Switches.
>>>
>>
>> _______________________________________________
>> evlatests mailing list
>> evlatests at listmgr.nrao.edu
>> https://listmgr.nrao.edu/mailman/listinfo/evlatests
>
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.nrao.edu
> https://listmgr.nrao.edu/mailman/listinfo/evlatests
More information about the evlatests
mailing list