[HamWAN PSDR] VOTE: Power backup expansion project

Bart Kus me at bartk.us
Fri Feb 5 20:31:32 PST 2021 

Answers in-line.

On 2/5/2021 4:46 PM, Tom Hayward wrote:
> The cost seems reasonable for an experiment at two sites, so my vote
> is Aye, but I have some questions...
>
> What is the grand design for our power architecture going forward?
There isn't a unified design.  We have different sites with different 
requirements and restrictions, so I would expect we'll use a variety of 
approaches.
> So far, we have a mix of AC and DC powered sites. UPSs are an obvious
> choice at sites that have full-size servers, but where we only have
> networking gear, a 24V battery bank is an option. We have done this at
> some sites. Are you advocating for standardizing on UPSs at all sites?
> Or just those with servers? Is this efficient? (Does efficiency even
> matter?)
For the sites in downtown Seattle, atop apartment buildings, we deployed 
giant battery banks and use DC feed.  This is a good match for these 
locations, since we can't be running generators at residential 
facilities, so batteries are the way to go.

For these mountain-top locations where we have limited 2-post rack space 
and no server load (yet), I think the 2 UPSes I purchased are a good 
fit.  If we add servers, we may want to expand those banks, which these 
units let us do.

Running on 120VAC gives us maximal equipment flexibility at the cost of 
some efficiency loss.  Answering this question actually made me do the 
efficiency science just now in the lab!  I connected the UPS to my 
"battery simulator" power supply, and removed 120VAC input. The no-load 
overhead draw is 77W-100W, depending on battery voltage (44V-56V).  
Adding a 450W load (small vacuum cleaner) to the UPS showed nearly 100% 
efficient conversion (10A draw @ 56V, including the 1.7A overhead from 
idle).  So it seems the only inefficiency is that static no-load 
overhead.  It's surprisingly high.
> Would it behoove us to use a redundant array of inexpensive,
> low-power, DC-powered (likely ARM-based) servers to extend runtime of
> our higher-layer networking services?
Maybe?  Although we can load-shed the servers if a site falls onto 
battery power.  Choosing servers is a more complex discussion.
> Do we plan to support non-HamWAN loads on these battery systems (e.g.,
> a DMR repeater in addition to the networking stack)?
I'm not sure if we should support non-HamWAN loads on these.  We did 
plug the VHF DMR @ Rattlesnake into the new UPS, however it is on its 
own outlet group.  If we get an alert about power, we can load-shed as 
appropriate.  Perhaps repeaters that want to share in the power backup 
can supply additional battery storage?

--Bart

> Tom
>
> On Fri, Feb 5, 2021 at 2:36 PM Bart Kus <me at bartk.us> wrote:
>> Responses in-line.
>>
>> On 2/5/2021 2:07 PM, Dale Skyllingstad wrote:
>>
>> I agree, this is a great idea.  Bart and I discussed it a little the other day.  I think our priorities should be the non-commercial/ham sites that have less reliable generators and therefore a greater likelihood of extended downtime.  Sites like Bawfaw that have public safety repeaters have great generator track records and should be at the bottom of our list, in my opinion.
>>
>> Agreed!
>>
>>    We may even be able to get away with smaller capacity units at these sites.
>>
>> I think we should still strive for 24 hours as a minimum in case a well maintained generator fails.  Possibly extend beyond 24 hours for known-flaky sites.
>>
>> Since we're using an on-line/double conversion UPS, dirty generator power shouldn't cause a failure.
>>
>> These units I bought are line-interactive, not double-conversion.  Double-conversion usually carries a price and availability premium.
>>
>>    Earlier this year, my gear at Baldi failed to come back on generator when my [standard] UPS didn't like the slow 50-something hertz the Wiztronics generator was producing.  An on-line unit shouldn't care, within reason.
>>
>> Yes, specifically for the Baldi Wiztronics shack(s) where we've had a switch get wiped due to generator failover, we should consider a double-conversion unit(s).
>>
>>
>> With the deployment of multiple UPS battery banks now, we should think about implementing a battery replacement schedule and budget.
>>
>> Absolutely.  For each of the UPSes in this email, the battery replacement cost is $262 (after shipping + taxes), and rated battery lifetime is 3-5 years.  I have saved the battery replacement date in the firmware.  I'd also like to see automation of calibration tests (runtime estimation), to alert us if the pack has degraded or if the load has increased.
>>
>> --Bart
>>
>> Dale
>> AH6ET
>>
>> On 2/5/2021 12:41 PM, Bart Kus wrote:
>>
>> Hello,
>>
>> During the recent wind storm event in January, HamWAN lost utility power to some sites.  This triggered network outages that affected users and repeaters that relied on those sites.  It was far from our finest hour.  "When all else fails", HamWAN should not be the first in line.  I'd like for 2021 to be a year we focus on reliability.  Having a reliable power system at every site seems like a good place to start that journey.  In each of the power failures in January, a generator system failed to activate.  In these cases, we need at least 24 hours to mount a response in fixing generators.  This means our battery backup systems should be sized for at least 24 hours of runtime.  I have acquired a couple UPSes that can achieve this.  They are composed of the main UPS:
>>
>> https://www.apc.com/shop/us/en/products/APC-Smart-UPS-XL-Modular-3000VA-120V-Rackmount-Tower/P-SUM3000RMXL2U
>>
>> And one external (and chainable!) battery expansion module:
>>
>> https://www.apc.com/shop/us/en/products/APC-Smart-UPS-XL-Modular-48V-Extended-Run-Battery-Pack/P-SUM48RMXLBP2U
>>
>> They also feature an AP9631 network interface module in each UPS, so we can monitor and control the power system.  As you can see, each UPS costs about $3600 new, so the two I acquired would be $7,200 new.  However, I managed to find used units and purchased brand new batteries for them.  The total cost for these two systems is $963.71.  One of them is deployed at Rattlesnake ("Snoqualmie shack") already:
>>
>> https://photos.app.goo.gl/HTomAzJUnrjuJ3bw8
>>
>> If we ever want to expand the battery capacity of these, we can simply plug in more battery modules.
>>
>> I'm hoping to deploy the 2nd system in the other building on Rattlesnake ("DNR shack"), but that hasn't happened yet.
>>
>> --Bart
>>
>>
>> _______________________________________________
>> PSDR mailing list
>> PSDR at hamwan.org
>> http://mail.hamwan.net/mailman/listinfo/psdr
>>
>>
>>
>> _______________________________________________
>> PSDR mailing list
>> PSDR at hamwan.org
>> http://mail.hamwan.net/mailman/listinfo/psdr
>>
>>
>> _______________________________________________
>> PSDR mailing list
>> PSDR at hamwan.org
>> http://mail.hamwan.net/mailman/listinfo/psdr
> _______________________________________________
> PSDR mailing list
> PSDR at hamwan.org
> http://mail.hamwan.net/mailman/listinfo/psdr

More information about the PSDR mailing list