The Ups & Downs of UPSs

I recently mentioned that I had the great misfortune of discerning the life of my UPS batteries, followed by replacing my UPS batteries. I replaced all the batteries, so I was given a chance to test them for run time, etc., without regards to keeping the UPS online.

To begin with, a lead-acid battery is an analog device. It's a device that converts chemical interaction into electrical power, though not at a linear rate. Chemistry is not my forté, so we will limit our discussion to the electrical side of the battery — with one quick side note. Vented batteries give off hydrogen gas. If you have a lot of batteries or UPSs in one area (especially an enclosed area), you need a hydrogen content alarm. If that content goes above 4%, you should turn off the power to all battery chargers. Outside venting should reduce the hydrogen content. The 4% point is when hydrogen in the air can explode. This is a bad thing.

Let's review some other types of battery testing. Capacity test uses a load bank with a constant current capability. Each cell in each battery must be monitored continuously. Problems can include the time needed; each battery could take up to 10 hours to test. You need an external load that matches the impedance of the live load. Reliability of the test is high if you truly test the battery's absolute capacity.

Resistance testing measures the internal resistance of each cell. As the battery ages, the resistance increases. Increased internal resistance indicates degrading or aging and can point to premature failure. This test can be done while the battery is in use. It is non-invasive but does not provide absolute capacity.

Let me offer a few bits of information regarding ohmic testing of batteries. Internal ohmic testing offers no conclusive proof, nor is it an accepted industry technique for determining electrical capacity. Ohmic readings should be used as a trending tool, but not as a substitution for actual capacity testing. Ohmic readings at the factory — AC impedance, AC conductance, resistance, or DC resistance testing — can be misleading and should not to be used as a baseline.

Another parameter worth noting is battery float, or the energy applied to the battery to maintain its charge. Temperature, probe placement, and differences in the impedance measurements all are major reasons factory readings maybe different from in-the-field readings. Ohmic values vary from manufacturer to manufacturer, as well as capacity and actual physical size. Float stabilization can take weeks or even months to establish.

The main three parameters that impact the accuracy of ohmic field testing include the baseline. This is the ohmic value of a perfectly good battery. All further tests are compared to the baseline. If there are errors in the field readings, the ohmic baseline may have no value as a reference. Error rates can be ~20% average or as high as ~37%. Whether a battery is good, OK, or failing, an important determination is the baseline and its accuracy.

Other issues are manufacturers' baseline setups. The battery tester will vary from model to model. The person in the field is heavily involved with the accuracy of the baseline, test results, and test repeatability. Remember that manufacturers want their products to meet their specifications and not have the piece returned for warranty replacement. Each manufacturer — whether making a battery, tester, or meter — needs to meet the published specifications, but how that test is performed may have more meaning. The technical staff in the field does not want to keep changing out batteries and constantly buying test equipment until one unit performs as expected. Maintaining uptime is the primary desired result, not developing a detailed test procedure for the field technicians.

Data analysis is the final variable. It has a substantial impact on ohmic testing reliability and any decisions made as a result of these tests. When these data points are combined, our procedures for ohmic testing start to take shape. But are we testing without technician errors and with proper procedures, correct probe placement, and reliability in measurements?

Factory readings using impedance meters from an ABC company and readings by a technician or engineer using a meter from GHI company will typically offer different results. The manufacturer's data is probably useless to the end user for creating a baseline. If you are concerned with actual battery capacity, stop using the internal life meter. Collect voltage, ohmic readings, and an accurate baseline that you create. Operating and testing temperatures are also key in data analysis. Don't forget to use the same calibrated meter and loads. This is what the state of health of the battery system takes.

Related posts:

30 comments on “The Ups & Downs of UPSs

  1. eafpres
    January 17, 2014

    Hi Peter–where does the age-old “test the electrolyte density” fall into your spectrum?  Many shops routinely test the electrolyte and advise customers on the state of the battery.

  2. samicksha
    January 17, 2014

    Your blog reminded of Sulfation, the deep discharge testing is itself damaging to batteries due to the chemicals in the discharged battery.

  3. PZman
    January 17, 2014

    I am a big fan of removing sulfation. I believe it is the main cause of battery failure followed by overcharging. There a few products that remove sulfation but I never tried them on sealed lead acid. Most sulfation occurs when water levels drop below top of plates. Sounds like an interesting test. Thanks.


  4. PZman
    January 17, 2014

    I have been taught that electrolyte density and specific gravity are alike. It is a simple test that tells how good the battery is, how the charge is. The specific gravity is an indicator of how the charge is dropping off but not a teat of battery life. If there could be a way to constantly monitor the SG maybe we got something here. As always thanks for everyone's comments.


  5. eafpres
    January 17, 2014

    @Peter–yes, specific gravity and density are the same for this purpose.  An online test would probably be feasible; wonder if anyone has done this.


  6. antedeluvian
    January 17, 2014

    There is a discussion on lead-acid batteris on the EE Times designlines forum here. I have cross posted.

  7. Davidled
    January 17, 2014

    Automotive Application The following items are reviewed: Static capacity testing Battery cycling Power performance Drive profile Safety guideline

  8. Netcrawl
    January 18, 2014

    @Peter that was great! thanks for a great post, this is exactly the problem with UPS, although sealed battery are sometimes referred to many as maintenance free they still require some scheduled maintenance, maintenence free refers to the facts that UPS do not require fluid, without service or regular maintenance our UPS may experience something like reduced protection and heat generating resistance at the terminals.      

  9. Victor Lorenzo
    January 18, 2014


    About ten years or more ago I collaborated with a colleague for making one data acquisition system (AFE+A/D+MCU) capable of meassuring signals (electrochemical noise) produced during steel corrosion in presence of acid and reading your article made me remember that project.

    Have you ever made or seen any study about the noise produced while discharging the battery under a constant load ('purely' resistive load)?

  10. PZman
    January 18, 2014

    Victor thanks for bringing up electrochemical noise, I remember some writings about it. I never thought about it in relantionship tp batteries. Should be interesting idea, I will do some research.

  11. Davidled
    January 19, 2014

    Electrochemical noise would be generated by thermal energy with oxidation and potential energy in the electrochemical cell. If this kind of noise would be detected by some type of sensor, it could be minimized as feedback output controls the reaction of electrochemistry.

  12. Victor Lorenzo
    January 19, 2014

    @DaeJ, the idea was to use electrochemical noise signals as one more health indicator by trying to identify kind of a pattern or patterns (waveform in the time-domain, spectral density/spectrogram in the frequency-domain or simply statistics) directly correlated with the processes that lead to battery aging.

  13. samicksha
    January 20, 2014

    I agree you PZman, restoration becomes more difficult the longer the battery is allowed to stay in a low SoC.

  14. samicksha
    January 20, 2014

    Seems valid fact, for a instance if a battery has been completely discharged and charged for few minutes, then during the short charging time it develops only a charge near the interface leading battery voltage rise to be close to the charger voltage so that the charging current decreases significantly. After a few hours this interface charge will spread to the volume of the electrode and electrolyte, leading to an interface charge so low that it may be insufficient to start the car.

  15. Davidled
    January 20, 2014

    Electrochemical Signal, not Noise

    I wonder if the application needs electrochemical noise signal to indicate health status of battery. If understanding correctly, if system requires electrochemical signal, not noise, it would need some kind of filtering (high, band, low pass filter) to remove noise before feeding into identification of pattern.

  16. Victor Lorenzo
    January 20, 2014

    It is called noise because of its mostly non stationary, stochastic  and unwanted nature. This kind of noise is mainly originated in processes that occur in the metal surface which is in direct contact with acid/base solution, it also has a high correlation with the magnitude and occurrence frequency of events like gas bubbles and metal oxidation.

    I wondered if this signals would be useful in determining the presence/depth of sulfation, hidrogen bubbles in electrode surfaces or even phisical electrode plates damages.

  17. Davidled
    January 20, 2014

    For Stochastic Processor with Probability, I recommend two books: “Probability, Random Variables, and Stochastic Processes,” is written by Athanasios Papoulis; “Probability Random Variables and Random Signal Principles,” is written by Peyton Z. Peebles, JR. They provide solid background knowledge for noise with random signal processing to engineer.

  18. eafpres
    January 21, 2014

    @Victor–” it also has a high correlation with the magnitude and occurrence frequency of events like gas bubbles and metal oxidation”  

    In many applications today, everyone is finding that getting more data is low cost (i.e., sensors are lower cost, data acquisition is lower cost, bandwidth to send the data is low cost) so they are looking at large “stochastic” data sets and trying to find “signatures” of certain events.  I am opposed to doing this purely as a mathematical excercise becuase you don't know where the noise and the signatures come from.  However, if you have a model, as you indicate, then this battery problem looking at noise spectrum (maybe doing FFT or multi-variable signature analysis etc.) could be very useful.  It stands to reason that if you have a model of a process, then a “real” process should be a “noisy” version of the ideal model.  You can then collect the signature of “normal” noise, and detect when it becomes abnormal.  Thus, you might derive an early warning of problems in the battery just by a simple data collection of frequency and amplitude (even though the main system is DC).

    Are you aware of any papers about this noise signature for lead acid batteries?

  19. Victor Lorenzo
    January 22, 2014


    I agree with you in your comments. Previous process model inference or determination plays a key role in systems analysis and supervision.

    I'm not aware of any paper covering this kind of application for lead acid batteries and it seems to me like an interesting topic. Most writings I've read about this subject cover studies focused on measuring signals produced during steel corrosion. Mostly because that was the main focus in the project I collaborated with.


  20. goafrit2
    January 22, 2014

    >> I recently mentioned that I had the great misfortune of discerning the life of my UPS batteries, followed by replacing my UPS batteries.

    Very curious @Peter, where do you live? UPS is for developing worLd and I am yet to see one in operation except in some big data centers but having it at home or small office is over-preparation. Light goes out say once per 4 months for about 1-2 minutes. Had to justify any investment on UPS.

  21. samicksha
    January 23, 2014

    We have been talking about batteries discharge and charge, any one remember Peukert's law, wherein he speaks about he capacity of a lead acid battery in terms of the rate at which it is discharged.

  22. Vishal Prajapati
    January 24, 2014

    Does anyone have FFT of electrochemical noise? What type of spectrum it covers? If battery can be diagnosed with such non invasive menthod, it would be great to see such application. Can it be significant to effect the portable analog electronics like medical equipments?


    Predicting life of battery has been a very complicated subject. I have tried it using gas gauging ICs available in the market. But no of chargin cycles remaining and amount of sulfation are a bit complecated areas.

  23. Buck-on-Bass
    January 26, 2014

    Dave Fader worked for Bell Labs and was involved in the development of the “Round Cell” for the telephone industry.  Several times he commented “There was no such thing as a maintenance-free battery.  It is a maintenance-resistant battery.”

    I read the post and the comments.  First, I don't know of anyone that used a float voltage of 2.32 volts per cell for a flooded cell at room temperature.  If you check with those making large flooded lead acid cells for stationary applications, you will find the typical cell voltages are 2.17 volts to 2.25 volts.  For longer service life,, lower cell voltages are used.  To get greater energy storage, higher cell voltages are used.  When a group of cells connected in series start having the individual cell voltages drift apart, the battery string is typically “equalized” at around 2.33 volts per cell.  Some users will equalize after a discharge to attempt to restore the full capacity to a battery as quickly as possible.

    Various alloy materials are incorporated into the battery grids to obtain some desired properties.  Calcium is used for standby batteries to reduce the float current and water usage.  Some call these batteries lead calicum batteries.  Instead of calcium, antimony is used in some batteries.  Antimony provides greater cycling capability but has higher float current and greater water usage.  Introducing other materials into batteries may adversely affect the batteries' performance.  This is why battery manufacturers provide information about what water is acceptable for adding to (watering) a battery.  This is why the noise analysis of the acid reacting to a metal plate may not be good for a continuous monitoring method.  

    The use of UPS systems on small computers or large datacenters and everything in between is an economic issue, and a convienence issue and not only a question of utility reliability.  If a $200 UPS prevents the loss for 4 hours of work a year from the lights blinking in a thunder storm or a momentary interruption from other causes, that UPS have more than paid for itself in my engineering group.   The UPS also provides transient protection which may prevent hardware failures in addition to data loss.  How expensive is a 10 minute outage on a mainframe system with 300 users not able to work while the system is restarted?  What about the data loss?  Maybe a $10,000 UPS is a viable way to avoid those interruptions.  If you are selling online services, what is the impact on customers?  Look at cloud service providers to see what interruptions do to their businesses.

    All the applications I have mentioned so far are considered to be stationary battery applications.  Batteries used for forklifts and similar applications are considered to be industrial battery applications.  In these applications the batteries discharged in the applications then recharged over several hours, typically 8 hours.  The objective is to quickly recharge the battery without an excessive temperature rise on the battery.  These batteries typically last 5 to 10 years and have a cycle life of 1000 to 2000 cycles.  The depth of discharge and the how long before recharging are a big influence on battery life.

    If a flooded battery is discharged and left in this state, it will sulfate.  The discharge may be by a user or by the battery' self-discharge rate.  This is why manufacturers have requirements for charging batteries in storage for long perions of time.  The battery manufaacturers have published information on how the maintain batteries in storage for months.

  24. goafrit2
    February 19, 2014

    Predicting life of battery has been a very complicated subject. I have tried it using gas gauging ICs available in the market

    Battery is a chemical product which is very complicated. Though it turns out that it is the basic tool for electrical engineers, there is nothing in battery that is electrical. This is chemistry at the highest level. My thinking is that you can accurately predict the life spam of a battery by looking at how the bonds in the chemicals degrade or break. So, it is chemistry that can accurately lead this prediction over circuits which end up becoming a burden on the DUT.

  25. Vishal Prajapati
    February 20, 2014

    The approach gas gauging ICs are using is probably just providing marginal estimation of battery life and its charge status. But as of now we don't have any concrete approach for looking in to battery chemistry, we are spared with this only. I don't know if there is any novel approach has been invented for this.


    If not this scenario needs complete makeover. The current approach needs to be revamped from scratch to have a proper estimation of battery capacity and its life.

  26. goafrit2
    March 22, 2014

    >> Does anyone have FFT of electrochemical noise? 

    I think the best strategy will be looking at the levels of atoms and molecules to see how bonds break. FFT will not offer much in my knowledge.

  27. goafrit2
    March 22, 2014

    >> The use of UPS systems on small computers or large datacenters and everything in between is an economic issue, and a convienence issue and not only a question of utility reliability.

    UPS adoption is very low. Not many companies use it. I know in the developing world with sporadic power supply, UPS is seen as one of those things everyone must have. But you do not expect this to be part of small company budgets in U.S. They think they can count and rely on the grid.

  28. goafrit2
    March 22, 2014

    There are many laws and knowledge base in battery technology. Yet, the problem is that many things are not known about materials, discharge and overall ways to improve efficiency. When a product is incubated by chemists for electrical engineers to deploy and use, you will notice a kind of lost knowledge base. Google papers on battery, they are everywhere. Yet, have an issue on battery design, you will notice not much is really known in some areas.

  29. goafrit2
    March 22, 2014

    >>   The UPS also provides transient protection which may prevent hardware failures in addition to data loss.  How expensive is a 10 minute outage on a mainframe system with 300 users not able to work while the system is restarted?

    You have explained why cloud computing is becoming very attractive. If you have a technology that caches so that you never lose data as you work, a 10min power failure will not be catastrophic. Most laptops now can take you to one hour.

  30. samicksha
    March 24, 2014

    You are right goafrit, Manufacturers supply run-time rating in minutes for packaged UPS systems. Larger systems such as for data centers require detailed calculation of the load and battery characteristics to ensure the required endurance is attained.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.