Silencing a 2U UPS

Arseni Mourzenko
Founder and lead developer
July 6, 2019
Tags: hardware 10 short 48

The following article explains an unauthorized modification of an electrical appliance which runs at high voltage and contains a lot of chemical stuff which doesn't play nicely at high temperatures. The manipulation explained in the article voids the warranty and modifies the product in a way which wasn't tested by the manufacturer. Unless you have enough experience, DO NOT ATTEMPT TO DO THE SAME MANIPULATION WITH A SAME OR SIMILAR DEVICE. Not only you'll void the warranty, but you would be at risk of permanently damaging the product or compromising your own safety.

I have a home of­fice with a bunch of servers and net­work hard­ware. In or­der to pro­tect my­self against oc­ca­sion­al prob­lems with the pow­er grid, I had two Eaton UPS units, but a firmware up­grade bricked both of them, so I had to buy an­oth­er one. I end­ed up pur­chas­ing Eaton 5PX2200RT.

This mod­el ap­peared to have one prob­lem: noise. It has two fans, Sunon ME80251VX-000U-A99, rat­ed 37.0 dBA. This might not seem ex­treme—af­ter all, most gamer's gigs have 40 dBA fans and high­er, and servers usu­al­ly run with 50 dBA or 60 dBA fans—but I made some ef­fort to en­sure all hard­ware is very qui­et, and it ap­peared that the UPS was the one which was mak­ing most of the noise. So it was time to re­place the fans.

The first chal­lenge was to find a qui­et fan which is good enough for the job. De­spite the de­vice be­ing 2U, there is no much place left in­side, which means that the fans should have a high sta­t­ic pres­sure. How­ev­er, as I ex­plained in my pre­vi­ous ar­ti­cle, there are no much qui­et fans with high sta­t­ic pres­sure out there, be­cause or­di­nary users don't care about sta­t­ic pres­sure, and users who buy fans for servers don't care about noise.

I fi­nal­ly de­cid­ed to get two Silent Wings 2 80mm BL060 fans. At 2.25 mm/H₂O, they have a lit­tle less than half the per­for­mance of the orig­i­nal ones rat­ed at 5.84 mm/H₂O. Prob­a­bly not a wise choice, since, ac­cord­ing to the man­u­al:

The 5PX mod­els that are cov­ered in this man­u­al are in­tend­ed for in­stal­la­tion in an en­vi­ron­ment with­in 0 to 40 °C [...].

In sum­mer, it is not un­usu­al to have the tem­per­a­ture in­side ris­ing up to 35 °C around the rack cab­i­net. At the same time, the UPS is very un­der­used for the mo­ment, and will be for the next few years. Nev­er­the­less, I'll mon­i­tor the tem­per­a­ture of the air flow ex­it­ing from the rear fan to en­sure it re­mains with­in lim­its.

In­stalling new fans

When the new fans ar­rived, I had a bad sur­prise. The con­nec­tors used by the orig­i­nal fans are very dif­fer­ent from the stan­dard 3-pin con­nec­tors. Hope­ful­ly, the volt­age was still the same, and in both cas­es, there is a black, red and yel­low ca­ble. A rewiring can be made by cut­ting the ca­bles of the fans and us­ing the old con­nec­tor with the new fan. I was glad enough to have a bunch of short volt­age re­duc­ing ca­bles which I used in­stead of cut­ting the ca­ble of the new fans; this way, I can still reuse the new fans with their orig­i­nal stan­dard con­nec­tors the day the UPS will die.

Fig­ure 1 The ca­ble mod­i­fi­ca­tion.

The orig­i­nal con­nec­tors are glued (the glue is still vis­i­ble on a 2-pin con­nec­tor on the left). I sup­pose that the man­u­fac­tur­er didn't ex­pect the fans to be changed, and es­pe­cial­ly not by some­one else than an au­tho­rized tech­ni­cian. Re­mov­ing the glue is sim­ple enough and doesn't leave any trace.

There is an ad­di­tion­al 3-pin con­nec­tor on the board which wasn't orig­i­nal­ly used. I'm not sure what is its pur­pose, since it doesn't look like there is a place (or a need for) a third fan.

Fig­ure 2 An ex­tra con­nec­tor.

Once the wiring is done, in­stalling the fans is very easy and straight­for­ward. Con­nect­ing them is easy as well, since the con­nec­tors are to­wards the mid­dle of the UPS, so there is no need for ex­tra long ca­bles.

Fig­ure 3 Front fan in place.

The fan on the back is close to a board, but it's still pos­si­ble to in­sert the fan with­out dam­ag­ing the board it­self if you take enough cau­tion.

Fig­ure 4 Rear fan in place.

Here's the over­all view. The fans are on the top left and top right.

Fig­ure 5 Over­all view.

Un­for­tu­nate­ly, I didn't think about mea­sur­ing the tem­per­a­ture with the old fans in or­der to be able to have some data about the way the new fans in­flu­enced the tem­per­a­ture. What I did, how­ev­er, is to test the UPS in an en­vi­ron­ment with a tem­per­a­ture of 29.8 °C, 62% RH. To do the test, I placed a SHT35 high pre­ci­sion tem­per­a­ture sen­sor in front of the rear fan with­out touch­ing the fin­ger guard, mea­sur­ing the tem­per­a­ture for sev­er­al min­utes first. The UPS load was 18% in av­er­age, re­port­ing 297-317 Watt. The probe re­port­ed a steady 32.9±0.1 °C for five min­utes.

Then, I un­plugged the UPS. A few sec­onds lat­er, I heard the noise of re­lays of the PDU, fol­lowed by a com­plete si­lence: the servers were down. When I plugged it back, it com­plained that the bat­tery was at 0%.

The next day, I did an­oth­er test. The room tem­per­a­ture was even high­er, at 30.5 °C, 54% RH. Same load as be­fore, 297-317 Watt. Mea­sur­ing ex­haust air showed a tem­per­a­ture of 33.8 °C. This time, when I un­plugged the UPS, the bat­tery start­ed to drew rapid­ly, and the pow­er would be lost with­in ten or twelve sec­onds (I con­nect­ed the UPS back to the grid just be­fore that; it was show­ing that the bat­tery is at 19%).

The day af­ter that, I re­peat­ed the test at 24.1 °C, 55% RH (same load). The ex­pect­ed time on bat­tery, ac­cord­ing to the UPS, was 27 min­utes, but with­in sec­onds, the bat­tery went from 100% to 97%, and I imag­ine that it would take it about two to four min­utes to drain out.

A few weeks lat­er, I per­formed a fourth test at 27.3 °C, 49.2 RH (same load). This time, it start­ed slow­ly de­creas­ing, about one per­cent every fif­teen sec­onds. This seemed promis­ing, be­cause fif­teen sec­onds times one hun­dred means twen­ty five min­utes—ex­act­ly the time es­ti­mat­ed by the UPS. How­ev­er, it went to 99, 98, 97, 96, and 95, and then sud­den­ly dropped to 20%. I sup­pose that it would shut down if I hadn't plugged it back im­me­di­ate­ly.

Based on those ob­ser­va­tions, I sup­pose that ei­ther I got a faulty UPS or an is­sue with the bat­ter­ies, or it doesn't en­joy too much my idea of silent, but less pow­er­ful fans. I would there­fore ad­vise any­one who de­cides to do the same ma­nip­u­la­tion to re­con­sid­er his choice.