E Q U I P M E N T N O T E B O O K

48 December 2002 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

By DAVID A. SELLERS, PEPortland Energy Conservation Inc.Portland, Ore.

Technology advances over thepast 20 to 30 years have resultedin increasing applications of ar-

tificial intelligence in building systems;hence the term “smart buildings.” Thesesame technology advances have providedus with a mechanism to become smarterabout buildings. The data logging and

handling capabilities of modern electron-ics can provide us with a wealth of operat-ing and diagnostic information in a small,integrated, cost-effective package to assistus in our learning process. The first stepin using this technology to our advantageis selecting the specific data-logging tech-nology that we will employ. This articlefocuses on how to go about selecting adata logger that meets the needs of yourproject and budget.

HARDWARE AND SOFTWAREWhen selecting a data logger, you will

be concerned with two general issues:hardware and software. From a hardwarestandpoint, you will be concerned with:

• Accuracy and repeatability.• Package size and weight.• Cost.• Real time display capability.• Time clock accuracy.

• Data format.• Memory.• Battery life.• Speed or sampling rate.• Starting modes (manually or auto-

matically initiated).• Input sensor compatibility.• Ease of set-up and deployment.• Ease of data retrieval and software

compatibility.• Ruggedness and robustness.From a software standpoint, when se-

lecting a data logger, you will be con-cerned with how the data set collected byyour logger is formatted for analysis anduse by other software packages. Most cur-rent technology loggers can provide thedata they retrieve in some sort of delim-ited file (see “Delimited File” sidebar),making it nearly universally useable byother software. A related software featureis the ability to retrieve the data set fromthe logger while it remains active in thefield. This will allow you to pull datafrom the equipment without having toremove it from the field and then rede-ploy it. Some equipment requires a com-puter to perform this operation, butother equipment can write data to flashmemory, floppy disks, or other mediawithout the intervention of a computer.

You might also be interested in whatsort of supporting diagnostic software isavailable to assist you with analysis of thedata you retrieve. Most manufacturers of-fer diagnostic software with whistles andbells that are designed to exploit the fea-tures of their particular equipment. Inde-pendent third parties also are starting toemerge with data-handling and diagnos-tic software packages that can work withdata from multiple manufacturers, eitherby importing it in some sort of delimitedformat or by interfacing with the data-base-management routines in the loggingequipment.

The sidebar “Resources for Sound Se-lection” describes two documents that of-fer guidance for the data-logger selectionprocess.

DATA POINTSRegardless of the hardware and soft-

ware package that you choose, an impor-tant aspect of any data-logging operationis the selection of the data points to bemonitored. For commissioning and diag-nostic purposes, the following pointsshould be considered.

Outdoor Conditions. Logging the out-door air temperature is nearly manda-tory. This is because outdoor air tempera-ture is a common denominator that cantie events together and provide a refer-ence framework when you are analyzinga problem. Adding outdoor air humidityto the data set can provide additional di-mensions beyond the two points loggedby allowing outdoor psychrometric con-

Selecting Data LoggersFundamental questions to ask before buying a system

Pho

to c

ourt

esy

of A

EC

David A. Sellers, PE, is a senior engineerwith PECI in Portland, Ore. He can bereached at dsellers@peci.org.

Delimited Files

Adelimited file is a file in which eachof the data values is separated

from each other by some sort of mark,such as a tab, comma, or space.Commas are one of the most commondelimiters. The files that use commasas delimiters are termed CSV files for“comma-separated value.” If a datalogger was monitoring outdoor air tem-perature and humidity once a minute,and you asked it to export the data ithad for the interval starting at midnightand ending at 12:05 a.m., the CSV file itgenerated might look like this:

DATE,TIME,TEMP,HUM11-01-02,12:00AM,45.2,8711-01-02,12:01AM,45.2,8711-01-02,12:02AM,45.3,8611-01-02,12:03AM,45.3,8611-01-02,12:04AM,45.4,86Commas separate each of the pa-

rameters, and a carriage returnseparates each time interval. Whensoftware capable of handling CSV filesimports the data and sees a comma, itknows that the next number it reads ispart of a new value that is not associ-ated with the preceding string ofnumbers.

Multi-input loggers interface with manytypes of sensing technology, onceconfigured to match the inputs selected.

49HPAC Engineering • December 2002

COMM

ISSIONINGOPERATION & M

AINTENANCEINSTALLATION

DESIGNditions such as enthalpy dew point to becalculated from the raw data using theASHRAE psychrometric equations.

Key process parameters.Diagnosticanalysis requires that process points otherthan those used to control the process bemonitored. For instance, many HVACsystems can be controlled most of thetime based on discharge air temperature.But if you are going to diagnose a prob-lem with the process occurring within theunit, you will need data from before andafter every heat and mass transfer elementto paint a complete picture and verifythat the discharge temperature control isworking efficiently and not wasting en-ergy by simultaneously heating and cool-ing. All may be required to truly under-stand if the discharge temperature controlfunction is working at peak efficiency andnot wasting energy by simultaneouslyheating or cooling the air stream.

Having more information may also al-low you to assess other parameters indi-rectly. For instance, an outdoor air per-centage may be implied based onconservation of mass and the outdoor air,return air, and mixed air temperatures.Bear in mind that it may not be necessaryto trend all of the process data simultane-ously although this can shorten theamount of time it will take to fully ana-lyze the system. But, if you have limiteddata logging capacity, then you can usu-ally break the data set down into logicalsubsets that match your logging capacityand provide the necessary information.

Key controlled variables.While inputsprovide a great deal of the necessary diag-nostic data required for the commission-ing process, the outputs from the controlsystem that impact the process can alsoprovide valuable insight into what is go-ing on. If an output can move somethingin the system or impact the flow througha heat-transfer element in the system,then it is bound to impact system per-formance. Being able to track the moti-vating input along with the system’s re-sponse to it can speed the verification anddiagnostic process.

CHOOSING AN INTERFACEInterfacing this data with a data logger

is really not that difficult, although theseemingly infinite variety of sensing tech-nologies available can seem overwhelm-ing when you first start looking intothem. Fortunately, most manufacturersoffer units with integral sensors that al-low the user to simply pick a logger thatwill be dedicated to the intended meas-urement function. Of course, this ap-proach sacrifices some flexibility whencompared to approaches using loggersthat can accept standard input signalsgenerated by transducers designed tomeasure nearly any physical parameterimaginable. Most manufacturers cansuggest a combination of logger and sen-sor technology that will meet the needsyou present to them. But if you are curi-ous and willing to spend a little time surf-ing the Web, you will find that there is awealth of information available to youthere, information that will aid you inmaking a more informed decision. Somegood starting points include:

• The DDC system input/outputtechnology tutorial on the Iowa EnergyCenter’s DDC Online Website(www.ddc-online.org) will provide youwith a good understanding of sensing,input and output technology used withcurrent technology control systems. Thistechnology is identical to the sensingtechnology required to interface data log-gers with the physical world.

• Watch PECI’s Website(www.peci.org) for notice of the publicrelease of the “Control System DesignGuide and Functional Testing Guide forAir Handling Systems: From the Fun-damentals to the Field.” This guide,funded by the California Energy Com-mission through a Public Interest En-ergy Research (PIER) grant, LawrenceBerkeley National Laboratory, and theU.S. Department of Energy, is in the fi-nal stages of development. The reportwill be released next year and will con-tain a chapter that takes a detailed lookat various sensing and actuating tech-nologies.

OTHER SELECTION OPTIONSPalm Pilots.Personal organizers such as

the Palm Pilot, are quickly making theirway into the data handling field. Somemanufacturers offer software that allowsthe data from the logger to be down-loaded for viewing. Others offer softwareand hardware packages that actually al-low the organizer to handle the data log-ging functions.

DDC system. If you are specifying aDDC system for your project, then,whether you know it or not, you are alsospecifying and selecting a data logger.Current technology DDC systems offerone of the most powerful but under-uti-lized data logging opportunities available.However, to be effective for trending anddata logging, the system needs to beproperly configured and programmed.

Using the control system to do yourtrending and data logging has numerousadvantages, including:

• You get more bang for your buck; in-stead of spending money to rent, deployand then retrieve data loggers. Thismoney is better spent to supplement theinfrastructure of the control system toprovide the desired data logging capacityfor the system.

• A control system configured to per-form data logging and trending will be amore robust system and generally exhibitsuperior, overall performance for day-to-day control functions.

• The trends that are set up in the sys-tem for startup and troubleshooting pur-poses can remain in the system as usefultools for the operators.

• Archiving the trend data can provideuseful troubleshooting information be-

Personalorganizers can bedata processorsand data loggers.(Photo courtesy ofthe Verteq)

E Q U I P M E N T N O T E B O O K

E Q U I P M E N T N O T E B O O K

50 December 2002 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

cause when a system starts acting funnyin real time, you can go back and com-pare what it is doing now to what it wasdoing the last time the same operatingconditions and triggers existed.

Most of the issues associated with suc-cessfully using the DDC system as a datalogger come back to design requirementsthat should be addressed in the project’sconstruction documents.

NETWORK ARCHITECTUREThe network architecture employed

for your DDC system is an importantfactor related to its performance as a con-trol system, but it is an absolutely criticalfactor in the context of using the systemfor trending and data logging. Currenttechnology controllers perform many oftheir functions in a stand-alone mode,thus poor network architecture will nothave a direct impact on their controlprocesses. But poor network architecturecan make the large volumes of data thatmust be handled to retrieve and archivetrend files very cumbersome and slow toaccomplish.

Fortunately, you don’t need to be an ITwizard to develop a control specificationthat will also give you good data loggingcapability with your DDC system. Mostof the major control companies have theability to do a reasonable job of trendingand understand the details and nuancesof their equipment and what it takes tomake their system perform in this man-ner. What they need from you is the nec-essary criteria in your construction docu-ments to allow them to provide thecapabilities you need in a competitivebidding environment. Addressing the fol-lowing issues in your specification will goa long way toward providing a useabledata handling system in addition to a bet-ter control system:

• Configure the system to minimizelow-level bus traffic due to trending oper-ations: Most control systems have severallevels of communication busses with thehigher level busses being cable of muchhigher data handling speeds than thelower level busses. By employing some ofthe specification techniques outlined in

the May 2001 “Control Freaks” columntitled “All Controllers Are Not CreatedEqual; Knowledge of the Differences isKey to Specification,” you can ensurethat the system you spec will meet yourneeds for data logging as well as control.The Iowa Energy Center DDC OnlineWebsite (www.ddc-online.org) is anothervaluable resource in this area.

Other networks have controllers atboth levels. This probably gives them anadvantage in terms of data handling ca-pability, but puts them at a cost disadvan-tage. In the context of the overall projectbudget, the cost differences are probablynot that significant, but, if you are a con-trol contractor trying to competitivelybid a project, they can be the differencebetween winning and losing.

• Consider specifying your trending re-quirements for each data point.This can beaccomplished by general statements to aspecific trending requirement for eachpoint that is called out as part of the pointlist. In either case, the goal is to define thefrequency and number of points thatmust be simultaneously trended withoutnoticeably impacting the usability andperformance of the control system.

CONTROLLER MEMORYThe amount of controller memory

can also have an impact on the system’sability to trend. “Too much” and “mem-ory” are mutually exclusive terms, be-cause you can’t go wrong in specifyingthat the controllers installed on yourproject be provided with the maximumamount of memory available. Memorythat is unused for programming purposesis available for trending. Having a largeamount of memory available for trending

minimizes the number of data transfersthat need to occur to move the informa-tion from memory to permanent storageon the system’s hard disk. This reducesnetwork traffic and helps keep the systemresponse time at low levels.

DATA RETRIEVAL AND ARCHIVINGNo matter how much memory you

put in your controllers, eventually it willfill up. If you are trending 16 data streamsonce per minute, you will generate 960bits of information to store every hour. Atthat rate, it won’t take long to fill upwhatever memory is available at the con-troller level. Most controllers allow youto determine what happens next. Thechoices are usually suspending trendingor over-writing the oldest data withnewer data. Either way, you will loosedata, which is undesirable in most in-stances. To circumvent this problem, thesystem needs to be programmed to peri-odically download the stored data fromthe controller’s memory to permanentstorage (usually a hard-drive in the opera-tors work station), thus making room forfresh data. Most systems can deal withthis issue, although for some it is a cus-tom programming problem while othershave canned routines that can deal withthe problem.

From a cost standpoint, having harddrives with capacities measured in tens ofgigabytes and read/write CDs, all avail-able for a cost of $200 or less, the need toconsider the cost of data storage is mini-mized. This is another area where yousimply need to specify what is needed tolevel the playing field and allow the con-tractors bidding your project to includeany costs in a competitive environment.The areas that will most likely be im-pacted by this are:

• The hardware supplied at the opera-tors work station to permanently archivethe data.

• The software supplied at the opera-tor’s work station for the control system“front end.” For some systems, the econ-omy version of their operator interfacesoftware will have a hard time supportingtrending, whereas the more sophisticated

Small dedicated-purpose, self-containedloggers minimize the time required to set-up and deploy them.

Pho

to c

ourt

esy

of M

icro

DA

Q

E Q U I P M E N T N O T E B O O K

51HPAC Engineering • December 2002

COMM

ISSIONINGOPERATION & M

AINTENANCEINSTALLATION

DESIGNversions will handle the problem as a mat-ter of course.

• Field time to set up the necessarytrends and archiving routines and verifythat they work.

TIME SERIES VS. COV TRENDINGOne technique that can be used to

maximize the use of available memory isto allow some of the data points you aretrending to only log data when there is achange of value (COV) rather than hav-ing every point write to memory once perinterval. The COV limits that you set fora point can also impact the amount ofmemory used by controlling how often apoint with a COV based record triggerwrites to memory. For instance if a tem-perature point is set to trend based onCOV with a COV limit of 5 F, then itwill only write a value to memory if thetemperature changes by 5 F from the pre-vious reading. If all other things are equala point with a 5-F COV limit will writeto memory much less often than a pointwith a 0.5-F COV limit.

However, this is a double-edged sword.Keeping the COV limit on the high sidewill preserve memory and minimize traf-fic on your network. But, if set too high,this may mask some serious performanceissues. For instance, if you specified a 5-FCOV limit for a discharge air tempera-ture sensor, you may not realize that thecontrol loop is hunting 4.5 F under someoperating conditions based on the datafrom the system. Thus, a problem thatcould be wasting energy and rapidlywearing out components would be unde-tected.

On the other hand, specifying a limitthat is too small can create communica-tion problems on the system by trying tolog to memory even the slightest varia-tion detected, be it real or false. For in-stance, if you set the COV limit for yourdischarge-air temperature point to 0.001-F, the system would attempt to logchanges that were most likely attributableto noise rather than real process changes.In extreme cases, this can crash the sys-tem, especially if the system is an oldersystem.

Based on experience, the followinggeneral rules will provide satisfactory re-sults and minimize problems:

• Trend analog values based on timerather than COV. Initially, you will wantto set the sample time at a very low value,perhaps a minute or so to avoid beingfooled by “aliasing.” Once you arethrough the commissioning process andhave gained some knowledge about thesystem, you may want to consider ex-panding the time frame to maximize theuse of memory and minimize networktraffic. If you take this step, as an opera-tor, you may still want to reduce the sam-ple time to a minimum on occasion, justto be sure that things really are stable.HVAC systems must operate under awide range of conditions, and it is notuncommon for a system that was stableduring the summer months to becometotally unstable during the swing season

or winter months or vice versa.• Trend digital values based on COV.

By their nature, digital parameters canonly have one of two possible values.Trending based on COV will capture anychange and minimize the amount ofmemory that is used. When using this ap-proach, check to be sure that the trend re-ports generated by the system will printCOV data mixed in with time series data.Otherwise, you may find it difficult toload analog and digital data into a diag-nostic software package or spreadsheetfor simultaneous analysis with analogdata. Being able to do this is quite helpfulwhen diagnosing operational problems.Most systems place a time stamp with theCOV information when they log it.

• Trend manually adjusted set pointsbased on COV and automatically ad-justed set points based on time.

Resources for Sound Selection

In 1999, the Portland Energy Conserva-tion Inc. (PECI) published a document

entitled “Portable Data Loggers—Diag-nostic Tools for Energy EfficientBuilding Operation” under a grantfunded by the EPA and DOE. Thisdocument is available for free as adownload by logging onto the PECIWebsite at www.peci.org.

The document addresses many ofthe important issues to consider whenselecting a portable data logger,including an evaluation checklist andcase studies, and is an excellentreference if you are considering pur-chasing such a device.

In May 2001, researchers atLawrence Berkeley National Laborato-ries published “Comparative Guide toEmerging Diagnostic Tools for LargeCommercial HVAC Systems.” The effortwas funded by the California EnergyCommission’s Public Interest EnergyResearch program and the U.S. Depart-ment of Energy and takes a detailedlook at many of the commercial andpublicly available diagnostic tools that

are emerging or currently available foruse. While many of the tools that arereviewed function by interfacing withthe database that exists inside com-mercial DDC control systems, some ofthem can also work with imported datasets which can be generated byportable data loggers. Examplesinclude:

• Enforma. A commercially availablesoftware package developed by a datalogger manufacturer.

• Universal Translator. A publiclyavailable software package developedby the Pacific Gas and Electric PacificEnergy Center that is currently comingout of Beta testing.

• UC Berkeley Fan Tools. A publiclyavailable software package that wasdeveloped by researchers at the UCBerkeley Center for EnvironmentalDesign Research.

This document is also available onthe PECI Website and is an excellentresource determining how to processthe data of a particular type or model ofdata logger.

88 January 2003 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

By DAVID A. SELLERS, PEPortland Energy Conservation Inc.Portland, Ore.

In the first article in this series, “Selec-tion of Data Loggers,” we reviewedwhat kinds of data loggers are avail-

able. In this article, we will describe someof the application considerations associ-ated with installing data loggers on yourproject.

HISTORYData loggers have been around for

quite some time in the process and scien-tific communities, perhaps for 25 or 30years. However it was not until the past 5to 10 years that the package size and cost

dropped to the point where they becameviable tools for the commissioning agent’sand HVAC system technician’s toolbox.

Evidence suggests that some sort ofdata logging technology has been em-ployed to allow us to document, analyze,and improve the performance of ourtechnology since man developed his firstsimple machines. Prior to the advent ofelectronics, the techniques employedtended to be simple, innovative, and ef-fective, but lacked the data-handling ca-pability and accuracy that are available to

us today. Examples of some of the olderdata-logging technologies include:

Operators with clipboards. Thismethod is still used in some plants to en-sure that key operating parameters areobserved, documented and responded toby the staff charged with keeping thingsin line.

Alarm clocks across starters. Wiring anelectric alarm clock across the coil in a120-vac starter control circuit was a sim-ple but effective means to gain insightinto the number of hours unattendedmachines operated over the course of ashift.

Low amperage fuses across contacts.Low amperage fuses wired across inter-

locking contacts provided a method ofdetection if a contact changed state whennobody was watching. The current flow-ing through the fuse after the contactopened in an effort to keep the load en-gaged quickly blew the fuse, leaving atell-tale indication for a troubleshooter.

Chart recorders. Circular chart andstrip chart recorders provided a way foroperating personnel to document variousoperating parameters over time. The for-mat is so widely accepted and easily un-derstood that several manufacturers offerelectronic recorders with displays thatemulate a strip chart or circular chart.

All of these approaches are still usefulin certain situations. However, moderndata-logging technology can providespeed, accuracy, and data-handling capa-bilities that technicians, engineers and

operating personnel could only dream of20 or 30 years ago.

No matter what process you are in-stalling your data loggers on, you need tobe aware of aliasing and adjust your sam-ple time accordingly. Aliasing is a phe-nomenon that results in the data set youcollect not matching what really is goingon. It occurs when the interval at whichyou sample data from a process is slowerthan a disturbance that is occurring in aprocess.

APPLICATIONS Data loggers can be some of the most

useful tools in an HVAC technician’s toolset. They can make a world of differencein the ability to troubleshoot and opti-mize older systems with stand-alonepneumatic, electric or electronic controlsand no form of large scale monitoring.In this sort of application they can:

• Provide enhanced troubleshooting ca-pability. One of the more difficult aspectsof troubleshooting HVAC problems istheir intermittent nature. Most problemsare difficult to pinpoint and solve without directly observing the event, espe-cially if they leave few permanent clues.By their very nature, HVAC systems op-erate under varied operating conditionsthat can trigger relatively complex inter-actions between various components andother systems. Without data logging ca-pabilities, a troubleshooting technicianresponding to a complaint related to anintermittent problem must attempt tosimulate the conditions believed to havetriggered the problem to be able to ob-serve and diagnose it or “camp out” withthe system until the event recurs. In ei-ther case, when the event occurs, thetechnician must be ready to observe anddocument multiple system parametersrapidly.

Strategically deployed data loggers al-low unattended documentation of multi-ple data streams as the system they servedetects and responds to various operatingsituations. This data can be retrieved and

Installation of Data Loggers“Aliasing” and other pre-installation considerations

The newest member of HPAC Engineer-ing’s Editorial Advisory Board (see p. 14),Dave Sellers , PE, is a senior engineer withPECI in Portland, Ore. He can be reachedat dsellers@peci.org.

No-risk trial before you buy

Imagine a future where you can go to a library and check out data-logging equipmentfor free. In at least one place, the future is now. The Pacific Gas and Electric Co has

created such a library, making a wide array of high quality, sophisticated instrumenta-tion available to anyone in their service territory. To find out more go towww.pge.com/003_save_energy/003c_edu_train/pec/003c1_pac_energy.shtml. Ifnothing else, looking through the on-line card catalog of tools will help you learn moreabout what is available in current data-logging technology.

E Q U I P M E N T N O T E B O O K

89HPAC Engineering • January 2003

COMM

ISSIONINGOPERATION & M

AINTENANCEINSTALLATION

DESIGN

analyzed by the technician subsequent tothe triggering event and without the pres-sures and potential errors associated withtrying to observe and analyze multipledata streams as they occur in real time.

• Support loop tuning. Data loggers that

can monitor, log, and display 2 or 3points in real time can be very usefulloop-tuning tools, especially if they in-corporate a display feature that allowstheir data to be displayed graphically inreal time. Not only do they allow the

controlled variable, process variable, andset point to be displayed and analyzed si-multaneously in real time as the loop istuned, they also document the loop-tun-ing process, including the results of openloop tests.

Data sampled every 15 min

Data sampled every 5 min

Data sampled every 3 min

Data sampled at 1 min

Data sampled at 1 sec

Real time

56

55 (set point)

54

56

55 (set point)

54

56

55 (set point)

54

56

55 (set point)

54

56

55 (set point)

54

56

55 (set point)

54

53

0 5 10 15 20 25 30Time in min

Tem

pera

ture

F

FIGURE 1. An illustration, using a manufactured data set, of the impact of sampling time on observed data vs what is really going on.When data is sampled every 15 min, everything looks fine. At every 5 min, a distorted wave form appears. Data sampled at every 3 minappears to be stable, but off-set from set point by about 1 F. The data sampled every minute appears to have a slightly lower peak andlower frequency than real time (bottom graph). The sampling rate of once per second reflects the real time data fairly accurately. Thereal time control system response shows a 2 F sinasoidal oscillation with a 3-min cycle time.

E Q U I P M E N T N O T E B O O K

• Spot check equipment operating pa-rameters for engineering studies. Fre-quently, the first step in a retrocommis-sioning process or some other sort ofengineering study involves a brief surveyof the building or system targeted for thework to evaluate if additional effort is, infact, warranted. Frequently, this effort in-

volves several hours to a day spent in thefield conducting a survey of the buildingsystems and equipment. Deploying miniloggers to monitor the operation oflights, computers, motors, and otherequipment coincidentally with the man-ual effort can provide a wealth of addi-tional information to help firm up the as-

sessment’s recommendations for very lit-tle effort on the part of the assessor.

DDC APPLICATIONSFor current-technology computer

based direct-digital-control (DDC) sys-tems, data loggers can provide a usefulsupplement to the monitoring capabili-ties provided by the control systems in-put and output points. In addition to thefunctions listed previously, loggers de-ployed in this type of application can:

• Provide data for troubleshootingprocesses not fully monitored by the DDCsystem. It is not uncommon for DDCsystems to be furnished with only thepoints necessary to control the HVACprocess with which they are associated.Frequently, the points required to controla process may not provide sufficient in-formation for diagnostic and/or func-tional testing purposes. Data loggers canfill this void by providing additional tem-porary monitoring capacity during trou-bleshooting and testing.

• Provide a faster data logging timeframe than is available from the DDC sys-tem. Even the best current technologycommercial DDC system tends to belimited to a data sampling frequency ofonce per minute or more. If you are deal-ing with a fast process or a fast distur-bance, aliasing can distort data or evenmask problems at this sampling speed.Figure 1 is a manufactured data set thatillustrates some of the effects of aliasing.Notice that the wave form produced by1-min samples is distorted from real timeto some extent. While this distortiondoes not hide the problem, it can changeyour interpretation of it in some cases.Thus, it is always good to remember thatreality may not be totally reflected inyour data sets. Taking a walk out to themachinery and verifying the electronicinformation using your built-in data log-ging equipment (your eyes, ears, and ex-perience) is often a good final step priorto taking action based on logged data. Itis also a good excuse to go “play” in thefield. You will just about always learnsomething useful to improve your engi-neering while you are there.

E Q U I P M E N T N O T E B O O K

90 January 2003 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

60 March 2003 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

By DAVID A. SELLERS, PEPortland Energy Conservation Inc.Portland, Ore.

In the past three articles in this series,we have reviewed how to select dataloggers and data-logging technology

and apply it to your HVAC systems. Oneof the biggest benefits of this technologyis as a commissioning tool. But, since thedata loggers are also machines, there are afew commissioning steps associated withgetting them up and running reliably andmaking use of their data.

DEPLOYING THE EQUIPMENTAfter going to all the effort to obtain

data-logging equipment and set up adata-logging plan, you are probably justitching to get the equipment out in thefield to gather information for you so itcan start paying for itself. Taking a fewminutes to pay attention to a few keypoints as you place the equipment in thefield will be well worth your while andcan make the difference between successon the first try and a meaningless orempty data set.

CHECKING POWER SUPPLIESMost portable data loggers use batter-

ies for power when they are deployed inthe field. For this reason, it is a good ideato get in the habit of checking and charg-ing the batteries in your loggers a day ortwo prior to the time you are actually go-ing to deploy them. In many cases, thecharge cycles are several hours to nearly aday, so discovering that your batteries aredead after you have flown to Poughkeep-sie for the day to deploy data loggers cancreate some problems, especially if yourloggers have built-in batteries. Loggerswith replaceable batteries may give you anout in such a situation, so carrying a fewspares is always a good idea.

SOFTWARE AND CONFIGURATIONSIt is also a good idea to verify the soft-

ware that you will be using to configurethe data loggers and retrieve informationfrom them prior to heading out into thefield. There could be some subtlety aboutthat new laptop you just purchased thatrenders useless the interface software youhave been using for years. Discoveringthis in the field on the only day you willbe on site to deploy the loggers can be abig problem. If you configure the loggersin your office a day or so before you willbe heading out with them, you provideyourself with some “wiggle room” if yourun into problems. This can be an espe-cially important consideration the firsttime you work with a logger, computer,or software that is new to you.

SENSOR INSTALLATION Before heading out to the field, it is a

good idea to think about where you willbe installing sensing elements and ifthose locations will require that you bringalong any special supplies. Things thatyou may want to carry with you as stan-dard items might include:

• Duct tape can come in quite handyfor securing sensors, cables and even dataloggers in temporary locations. It can alsobe used to patch insulation if you have tocut into it to pick up a reading.

• Hole plugs are useful for sealingopenings in ductwork after you have re-moved your sensor and will win youpoints with owners and operators whoare looking for signs of your professional-ism and/or are concerned with the in-tegrity of the systems. Sometimes, littlethings make big, long lasting impressionsas can be seen from the story in the side-bar, “On Being Prepared.” Plugs come ina variety of shapes and sizes for differentapplications and are available from a vari-ety of sources.

For high-pressure applications, youmay want “official” test ports from asheet-metal supply house.

For less-demanding, low-pressure ap-

plications, rubber stoppers or plastic-tube plugs from a science supply housewill suffice. Don’t forget to bring a porta-ble drill.

• Silicon caulk is also useful for sealingand vapor proofing holes you may haveto create in insulation to install yourprobe, as well as repairing the holes afterthe probe is removed.

• Quick-ties or tie-wraps are anotherhandy item to have in your tool bag forsecuring things temporarily. If you need along one, just string several shorter onestogether.

• Heat-transfer grease can come inhandy to improve the thermal responseof wall-mounted sensors and improve theapproach of surface-mounted sensors. Itcan be a little messy to work with, so youmay also want to carry a few rags with

Commissioning Data LoggersCaulk, duct-tape, flashlights, and other essential tools

E Q U I P M E N T N O T E B O O K

David A. Sellers, PE, a member of theHPAC Engineering Editorial AdvisoryBoard, can be reached at dsellers@peci.org.

On BeingPrepared

Aproblem came up on one of mylong-term projects and I was

unable to get to it immediately due toother commitments. So, Phil, one of mymentors, who was also familiar withthe project and happened to be in thearea for a meeting on another project,stopped by the site to take a look . Oneof the mechanics whom I had workedwith for years walked him out to theproblem area and, as Phil climbed upthe ladder to take a look above theceiling, he asked if he could borrow aflashlight. As the mechanic handed himflashlight, his only comment was “Davealways brought his own flashlight”.Phil’s expertise and quick solution tothe problem rapidly gained the respectof the operating staff, but for that onemoment, the fact that I had shown upwith the right tool when I was on site totroubleshoot made more of an impres-sion than Phil showing up with 10 timesthe knowledge and experience that Ihad at the time.

61HPAC Engineering • March 2003

you to wipe your hands subsequent to theinstallation.

• DC-voltage data loggers set up forsome of the more common input volt-ages, such as 2 to 10 VDC or 1 to 5 VDC

can provide a convenient way to piggy-back onto the signals coming into an ex-isting DDC system.

This can be particularly helpful if yourun into a signal you need to monitor butdon’t have the appropriate sensing ele-ment for.

Remember that most current loopscan be easily converted to voltage by sim-ply running the signal through a preci-sion resistor. So, carrying a few 250 and500 Ω precision resistors can be quitehandy. One caution when using this ap-proach is that it’s a good idea to run theconnection you plan to make past some-one familiar with the wiring practices forthe system you are working with to besure that you don’t cause problems withtheir I/O boards. (To learn more abouthow a current loop works, there is a good

application note from Scan Data atwww.scan-data.com/app-1115.pdf.)

Once you are out in the field and readyto install your sensors, there are severalthings to keep in mind:

• Many of the ducts and pipes that youwill need to monitor will be insulated, re-quiring you somehow penetrate the insu-lation to get a good reading. From asafety standpoint, if the insulation is as-bestos, you should not disturb it. At thispoint in most facilities, people knowwhere they have asbestos, but if there isany question in your mind, then it is bestnot to disturb it until you have assurancesthat it is not dangerous to your health.

Sometimes, you cannot make a hole ina duct or there is not a well in the pipewhere you need to measure. All hope isnot lost, however. If you insert the sensorunder the insulation so it is in intimatecontact with the surface of the pipe orduct, the reading you obtain will be veryclose to the actual line temperature, espe-cially if you take a minute to clean the

pipe and place the sensor in a dab of heattransfer grease. (Minco’s Application Aid16 on sensing fluid temperatures withthermal ribbons contains some interest-ing and useful information regarding sur-face-temperature measurement. Visitwww.minco.com/pdf/aa16.pdf.)

When using this approach, it is impor-tant that you cover the sensor with insu-lation so that the impacts of the ambienttemperatures are minimized. If the line iscold and could condense moisture out ofthe local environment, then vapor sealingyour penetration with silicon caulk is im-portant to keep the insulation from be-coming water logged, which will ruin itand also ruin your measurement sincethe water will interfere with the surface-to-surface contact between your probeand the line.

Remember, too, that intimate contactwith the line is important for the temper-ature probe, but can be a disaster for thecable attached to it. Lines with steam, hotwater, or hot air can quickly melt the in-

DESIGN

E Q U I P M E N T N O T E B O O K

Circle 178

COMM

ISSIONINGOPERATION & M

AINTENANCEINSTALLATION

Circle 181

sulation on the cable if it is lying againstthem, ruining your probe assembly anddata set.

• Even though a sensor installation istemporary, it is important that it not betoo temporary. For instance, the sensorthat you dangled through a hole in theduct at the fan inlet could get tangled upin the fan drive system. Generally, this isnot good for the sensor, the data logger,the drive, or your ego (not that anythinglike that has ever happened to me).

• It is also important to make sure thatyou position your sensors so they pick upthe data you want in a meaningful man-ner. For example, if you are installing atemperature sensor near a tee, mixingvalve or mixing damper in a pipe or duct,you may want to double check that youhave selected the appropriate branch, es-pecially if the tee is mixing two differentfluid streams. On electrical gear, it is im-portant to be sure that the signal at thepoint you are monitoring is compatiblewith your sensor and will give you a

meaningful data stream. For instance, in-stalling a CT on the conductors leaving avariable-speed drive may not provide use-ful data because at that point, the voltage,current, and frequency have been manip-ulated by the drive to control the motorspeed, and most CTs are intended for usewith 60 cycle alternating current.

CHECKING RELATIVE CALIBRATIONIf the parameters you will be measur-

ing will be subtracted from each other todocument the performance of a heat-transfer element or a prime mover (tem-perature difference, pressure difference,etc.), then it is a good idea to place all ofthe probes together and log 10 or 15 minof data with them all referenced to thesame value. This will give you a good ideaof the relative accuracy of the sensors toeach other. If there are differences (thereprobably will be), you will have docu-mented them and can use a formula inyour analysis spreadsheet to make any ap-propriate corrections. If you retain this

data as a part of your data set, you will al-ways have the correction factors availableto you. If absolute accuracy is important,you may also want to note the sensorreadings relative to a calibration standardthat is subjected to the same condition asthe sensors during relative calibration.

CONCLUSION: TURN ON THE DATALOGGER

This may seem obvious, but youwould be amazed by how easy it is to for-get to turn on a data logger. With someloggers, you simply have to trust that theyare running—there are no real indicators.But many of them have some sort of indi-cation, ranging from a flashing light to anactual display of the most recent datastring. Taking a few minutes to make surethat there is really something there cansave you weeks of time and ensure thatyou capture an important, non-reoccur-ring event.

For previous Equipment Notebook articles, visit www.hpac.com.

E Q U I P M E N T N O T E B O O K

62 March 2003 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

50 February 2003 • HPAC Engineering

COM

MIS

SION

ING

OPER

ATIO

N &

MAI

NTEN

ANCE

INST

ALLA

TION

DESIG

N

By DAVID A. SELLERS, PEPortland Energy Conservation Inc.Portland, Ore.

The data provided to you by yourdata-logging equipment canprovide a wealth of insight into

what is really going on in your HVACsystems. It is not uncommon for trenddata to reveal myriad problems in situa-tions where the more traditional indica-tors (occupant comfort, sound, etc.)would indicate everything is fine.

While the software packages discussedin the first two parts of this series can helpyou analyze your data to reveal a lot ofuseful information about what is goingon with your systems, modern spread-sheets can provide you with quite a bit ofinsight for little or no first cost since youprobably already have them installed onyour computer.

PRESENTATION IS IMPORTANTThe primary analysis technique when

using spreadsheets is to look at the data ina graphical format. There are a variety offormats for this. For example, you couldlook at a data set as time series data or as ascatter plot contrasting different parame-ters. Often, it is best to look at the samedata set in different presentation formatsbecause the different formats will tell youdifferent things.

The plots in Figures 1a & b are differ-ent views of the same data set. Both weregenerated from data loaded into a spread-sheet. The one on the top is a more con-ventional time series plot of an air han-dling system’s outdoor air, return air andmixed air temperature. This type of pres-entation is good for investigating:

• Instability. Hunting control loopsquickly show up as very “squiggly” lines.Stable loops show up as straight or gradu-ally changing lines with a narrow span be-

tween peak and valley. The data in figure1 indicate a reasonably stable system butalso reveal that the outdoor air sensormay be influenced slightly by somethingsince it is unlikely that the outdoor airtemperature actually varied by 3 or 4 F ina matter of minutes as the “squiggly” lineseems to indicate.

• Inter-relationships. Plotting datastreams simultaneously in this formatwill allow you to understand which eventtriggered what. For example, assume thatthe system under analysis had an inte-grated economizer cycle. At the outdoorair temperatures that were occurringwhen the data set in Figure 1 was ob-tained, one would anticipate that when

the economizer was active, the mixed airtemperature would nearly match the out-door air temperature because the out-door air temperature was above the re-quired discharge temperature for thesystem and the economizer would bedriven to the 100 percent outdoor air po-sition. If the economizer was disabled,the mixed air temperature would beslightly below the return air temperaturebecause mixing the cooler minimumoutdoor air quantity with the return airstream would lower the mixed tempera-ture slightly. When the system was off,one would expect the mixed air tempera-ture to nearly match the return air tem-perature as temperatures in the unit

Datalogger Operation TipsWorking with data: trend analysis and spreadsheeting

The newest member of HPAC Engineer-ing’s Editorial Advisory Board, David A.Sellers, PE, is a senior engineer with PECI.He can be reached at dsellers@peci.org.

FIGURES 1a (top) and 1b. Different views of an identical temperature data set from an air-handling system. While not immediately obvious, the manipulations associated withFigure 1b, which were developed by researchers at some of the National labs and a data-logger manufacturer, can quickly reveal a lot about how an economizer is working.

E Q U I P M E N T N O T E B O O K

HPAC Engineering • February 2003

COMM

ISSIONINGOPERATION & M

AINTENANCEINSTALLATION

DESIGN

equalized. When the economizerchanged over, there would be a suddenshift in the mixed air temperature.

The data set indicates that there seemsto be no particular correlation betweenthe economizer change over and the out-door air temperature, and in fact, itturned out that the operators were manu-ally controlling this changeover ratherthan allowing the automation system tohandle it. The data set also reveals that therelative calibration of the sensors is rea-sonably good because the mixed-air tem-perature sets on top of the outdoor-airtemperature when the system is probablyusing 100 percent outdoor air and the re-turn-air temperature and mixed-air tem-perature nearly match during what ap-pear to be the off cycles.

• Sequence of events. Sudden changes inmixed-air temperature most likely indi-cate that the economizer was enabled ordisabled or that the system shut down. Inthis particular instance, changes early andlate in the day correlated with the sched-uled operation of the unit and changes inbetween correlated with the operatorsmaking the decision to enable or disablethe economizer function.

The scatter plot in Figure 1b allows fora quick assessment of the general per-formance of the economizer it represents,once you understand the data patterns.Specifically, the plot is of the differencebetween mixed-air temperature and re-turn-air temperature vs. the difference be-tween outdoor-air temperature and re-turn-air temperature. The solid lines aretheoretical operating lines that represent:

• The line generated by a system oper-ating on 100-percent outdoor air (yel-low).

• The line generated by a system oper-ating with minimum outdoor air (red).

• The line generated by a perfect sys-tem with perfect sensors operating withthe setpoints associated with the econo-mizer under study; i.e. the economizer’stheoretical operating line (green).

Points generated by the real data arethen plotted against these reference lines.If the economizer was working perfectly,then the data pattern would tend to gen-erate a cloud around the green line.Clouds that are not on the line will have

characteristic shapes and locations rela-tive to the axes that will tell you some-thing. For instance, the cloud that isforming along the negative part of the Xaxis is generated by the temperatures cre-ated in the system when it is off with thehot water valve only opening as necessaryto prevent the mixed air plenum temper-ature from dropping below 40 F (as de-sired by the required control sequence).If the hot-water valve were kicked wideopen any time the system was off, as is thecase for some systems, then you wouldstill get a cloud along the X axis, but itwould shift to the left with the magni-tude of the shift controlled by the tem-perature generated in the mixed airplenum in that mode. A system that op-erates 24 hr per day would not exhibitthis cloud. Thus, a data set from a systemthat was supposed to be scheduled butdidn’t reveal this cloud, might cause youto question if the scheduled operationwere really occurring.

Note that to generate a cloud that cov-ered the entire green line, you wouldneed a data set that covered the entire op-erating season. The data set associatedwith this particular plot is 10 days worthof 1-min data. Note also that for dataplotted in this format, the sampling in-terval is not nearly as critical as it is for theother format. The points on this plot willfall on the line if the measured conditionsare what they should be at the instantmeasured, regardless of what was goingon immediately before or after that in-stant.

"PRIMING" YOUR DATA-COLLECTIONEQUIPMENT

No matter what software package isused, there are few things to be done upfront to allow the retrieved data to bemore easily digested for analysis, espe-cially if an EMS is being used as the datalogger. Some of this relates to specifyingthe correct performance parameters inthe first place, as has been discussed inpreceding articles in this series. But muchof it also relates to telling the controltechnician specifically how you will beusing the data to allow him or her to setup the trending package appropriately.

Most control systems offer a variety of

options to define how trend data is re-ported out of the system. From the con-trol system’s standpoint, the trend data isjust a bunch of numbers, probably hexa-decimal numbers or ones and zeros, nei-ther of which is very easy to interpret.But by setting up the trend reporting pa-rameters properly, you can get the controlsystem to provide this data in a meaning-ful format. For example, it is generallyeasier for spreadsheets and analysis pro-grams to deal with data presented in ver-tical columns with a time stamp—basi-cally a table of values with the pointnames across the top and the date andtime down the left-hand side. This is incontrast to a single long string where all ofthe times and values for the first parame-ter are presented, followed by the secondand then the third, etc. From a specifica-tion standpoint, what matters is that thecontractor is informed that you want tobe able to define how the data is pre-sented along with some key features, suchas the tabular format describe above.This will allow the contractor to buildsome time into the price to work withyou to get the exact format you wantwhen programming the system. At aminimum, you will probably want thesystem to be able to:

• Export data in a delimited format asdiscussed in the previous articles.

• Maximize the number of parameterspresented in each report. In the tabularformat, this would be analogous to maxi-mizing the number of columns in thetable.

• Maximize the number of samplespresented in each report. For example,maximizing the number of rows in thetable in a tabular format.

CONCLUSIONBy setting specifications up properly

and then coordinating with the controltechnician before they set up the trendingyou requested, you can ensure that thedelimited data files that will be providedfor analysis will be in a format that is theeasiest to work with. This will minimizethe amount of time you need to spendmanipulating the data and maximize theamount of time you get to spend havingfun.

E Q U I P M E N T N O T E B O O K

51

105HPAC Engineering • September 2003

A member of HPAC Engineering’s Editorial Advisory Board, David A. Sellers, PE, is a senior engineer special-izing in commissioning and energy efficiency. Over the course of his career, he has worked in the design, mechan-ical- and controls-contracting, and facilities-engineering fields in the commercial, institutional, and industrial-buildings sectors. He can be contacted at dsellers@peci.org.

A picture is worth at least a thousand

words when working with BAS data

Visualize BAS DataMicrosoft Works, although some of the more pow-erful formatting functions will probably not beavailable. These techniques can also be used on de-limited files exported from traditional data loggers.In fact, you can use a spreadsheet program to com-bine data obtained from data loggers, the EMS, and

even the National WeatherService (see the sidebar “Sup-plementing HVAC Data withWeather Data”) and othersources to further enhance

your analysis.

Many new HVAC systems come witha powerful built-in data logger in theform of the DDC system, which

may also be referred to as the Building AutomationSystem (BAS) or the Energy Management andControl System (EMCS).When properly config-ured, these systems can providean abundance of performancedata. Many systems have soft-ware modules available that en-able graphic analysis of thetrended data. But even if you do not have the soft-ware module that enables this function under theumbrella of the control system soft-ware, virtually all current technologysystems (and many older systems) willallow you to export the trend data in adelimited file format,1 which can thenbe imported into a standard spread-sheet program for analysis.

In most instances, you will still needto manipulate the delimited data files tomake them more readily analyzed in aspreadsheet program. After you do thisfor a while, you will discover that thereare a few tricks you can use to makeyour work easier. This article describe afew that I have used in the course of mywork. I have used all of the techniquesdescribed below with Excel (97 andlater versions) and have used many ofthem with Lotus prior to working inExcel. Many of them will also work inthe spreadsheet program contained in

By DAVID A. SELLERS, PEPortland Energy Conservation Inc.

Portland, Ore.

PHOTO 1.Don't forgetto select"All Files(*.*)" whenyou go toopen yourCSV file forthe firsttime.

OPENING THE CSV FILE IN ASPREADSHEET PROGRAM

For many, it would seem that the firstobstacle in this process is to get thecomma -separated-value (CSV) file fromthe data logger or control system to openup in the spreadsheet. Five or six yearsago, that could be a little tricky, requiringseveral steps including “parsing” data, aprocess that scanned the data file, foundthe commas or other delimiters, and cre-ated cells based on what it found. Recentversions of Excel will allow you to simplyuse the “Open” command on the “File”tab to select the file from your hard disc ora floppy.2 The spreadsheet will then use astring of questions in a dialog box to allowyou to control how the data is separatedinto cells; but, in my experience, if yousimply accept the defaults offered by thesoftware, you will create a usable spread-sheet file.

For me, the biggest trick during thisstep of the process is to remember to selectthe “All Files (*.*)” option when the“Open” dialog box appears (see Figure 1).Otherwise, you won’t be able to see yourfile. Remember, when you start, it is rawdata in the form of a CSV, and probablyhas a file extension like .CSV or .TXT.On most systems, the default file type se-lected by Excel in its “Open” dialog box isfor Excel file types, i.e., files with exten-sions like .XLS, .XLT, etc. and thus, onlyfiles with those extensions will be dis-played.

After you have opened the file, you willwant to save it as a spreadsheet file so thatyou can take advantage of the formulaand graphing features available in the

spreadsheetformat. InExcel, you ac-complish thisby simply se-lecting the“Save As” op-tion on the

“File” tab andentering thename of your

choice.

SORTING DATA TO ELIMINATE THETHINGS YOU DON'T NEED

Once you have opened your CSV fileand turned it into a spreadsheet file, youprobably will need to go through a fewquick steps to clean things up so that thegraphing can be accomplished morereadily. Usually, there will be things likeheader lines for every page and point def-initions that, while useful when viewingthe data as text, get in the way of plottingit. There will probably also be some holesin the data where the data collectionhardware simply didn’t capture a value.

The sorting features of spreadsheetsprovide a way to easily group and elimi-nate these sorts of things. For example, ifyou simply sort the spreadsheet based onthe first column in ascending order, thesort will pull all of the header lines to-gether at the top followed by the time se-ries data in ascending order. Be careful tounhide everything and select all of thecells before you do this, otherwise youcan get some strange results.3 Similartechniques can be used to pull empty cellsand rows together for elimination or rowsthat have things like “NO DATA” inthem. If you are doing scatter plots, elim-inating these data sets will not really im-pact the graph you get if you format thedata series as lines. If you are using dots,you will have gaps in the patterns createdby the scatter plot where you eliminatedthe data.

HANDING MISSING DATAMany spread sheet programs allow you

to control what the graphing functiondoes with missing data under the generaloptions setting. In Excel, this is under the“Tools” tab and the choices are to notplot the data, to plot it as zero, or to in-terpolate between values.

Excel will also allow you to add a trendline to a data set. This technique can alsobe used to create a solid line from a dataset with holes or to project what mighthappen based on what you know. If youwant to be really clever, you can formatthe data series color to match the chartbackground and it disappears (except forwhere it crosses grid lines or other data se-ries) and the trend line remains as the vis-ible indicator.

GETTING AT DATE/TIME COLUMN FROMRAW DATA

It is not unusual for the exported file tocontain the date/time stamp as two sepa-rate values, date and time. Some systemsbreak this down even further so that youend up with a column that representseach parameter (year, month, day, hour,and minute). Figure 2 is an example ofthe raw data from such a system. Whenyou graph the data, it is usually best tohave the date and time together in onecolumn because many spreadsheet pro-grams treat date and time as a serial num-ber. The most common bases are January1, 1900 or January 2, 1904. These datesare represented as one and are then incre-mented by 1 for each day there-after. Inthis system, 1 hour is 1/24, 1 minute is1/(24 × 60), etc. which is helpful to re-member when you are trying to scale thetime and date axis of a time series graph.4

In any case, if you initially have twocolumns, one for date and one for time,you can generate the desired date plustime column by simply inserting a col-umn into the spread sheet and then usinga formula in each cell of the new columnto add the value of the date and timecolumns for that sample together. Datasets that break the date and time downinto individual components, like those inFigure 2, require a little more effort to

106 September 2003 • HPAC Engineering

E N E R G Y - E F F I C I E N T H O T E L S

Column B C D E F G

Row Month Day Year Hour Minute Value

3 1 3 3 22 05 55.3

4 1 3 3 22 05 55.2

5 1 3 3 22 05 55.3

6 1 3 3 22 05 55.4

FIGURE 2. Illustration of a spread sheet with raw data set from a systemthat breaks the time and date down into separate values for eachparameter Figure 3 illustrates this process for the data in Figure 2.

make the conversion from integers to aserial number that represents the date andtime.

At first, placing formulas in all of thecells created by the inserted columns canseem like a formidable task, especially ifthere are thousands of data points. But re-member, all you really have to do is createthe formulas in the cells for one row. Youcan then copy and paste them to the re-maining cells in a matter of seconds via acouple clicks of the mouse.

USING THE TRANSPOSE FUNCTIONSometime DDC systems only print

the point names at the top of the data setand the labels for the data columns aregeneric such as point 1, point 2, etc.While this is not the end of the world, itwould be nice to be able to get the actualpoint names at the top of the columns sothat they show up as labels in yourgraphs. The transpose function allowsyou to easily do this an is illustrated inFigure 4. @

The key things to remember when us-ing this function is that it is an array func-tion. So, to use it, you first highlight thecells you want to transpose the data into(it should be the same number of cells asthe source of your data). Then you typein the transpose formula. However,rather than simply pressing enter afteryou are done, you need to press CTRL +SHIFT + ENTER. When you do this,the curly brackets will be added to your

formula, it will be placed in all of the cellsin the array you selected and the resultswill show up.

SCATTER PLOTS FOR TIME SERIES DATAScatter plots are a good way to plot

time series data because they allow you touse different time parameters for the xaxis, thus you can pull data from multiplefiles together into one graph. It also solvessome problems associated with missingdata or blank cells. You can still get linesinstead of little dots or squares by playingwith the formatting parameters associ-ated with each data series. Lines are usefulif you are trying to present things as atime series, i.e., this happened then thisand then this. Individual dots or pointsare useful when you are trying to presentthe data in terms of number of occur-rences of one condition when anothercondition was present. For example, plot-ting mixed-air temperature minus returnair temperature against outdoor tempera-ture minus return-air temperature to lookfor patterns (See Figure 1b in “DataloggerOperation Tips” 5

USING THE "CHART WIZARD"The chart wizard feature provided with

Excel frequently will provide a quick wayto take a first look at your data set. Inmost cases, if you can set up your data filewith the time/date information in theleft-most column and the point names inthe top row, you can generate a graph by:

1) Highlighting the data set, (includ-ing the time/date column and pointname rows).

2) Clicking on the Chart Wizard Icon(it’s the one that looks like a little 3-D bargraph).

3) Selecting one of the scatter plot op-tions.

4) Selecting the “Data in Columns”option on the data range tab.

5) Selecting “Finish”What you should get is a basic graph

containing all of the data you high-lighted. You can now use this as a sort ofbase-line graph to develop all of the othergraphs by copying it and then adding ti-tles, deleting data series, changing thescale of the axes, etc. Once you have gonethrough the process a couple of times,you will find that you can go from dataset to this base-line graph with a few sim-ple key clicks.

Often, what you will see in the base-line graph can guide you on where youwant to focus your attention. For in-stances, a set of wildly oscillating linesmay tell you that you need to focus in onthose parameters at that time frame to seeif there is an unstable control loop. Thiscan be accomplished by saving the graphto a new sheet and then eliminating theseries that are not of interest, rescaling thetime axis to only cover the range of timeof interest and then “tweaking” the otherchart options and parameters like the axisscale, colors, line weights, titles, etc. to

107HPAC Engineering • September 2003

E N E R G Y - E F F I C I E N T H O T E L S

Column B C D E F G

Row Month Day Year Hour Minute Month, day,and year

combined

3 1 3 3 22 05 1/3/03

1/3/03

1/3/03

Formulas

3 3

22

22 07

5 1

6

7

1

3 3 22 06

H I J K L

Hour andminute

combined

Serialnumber that

represents date

Serialnumber that

represents time

Serial numbersadded and formatted

as a date and time

Value

22:05 37624.00 0.920139 1/3/03 10:05 pm 55.3

22:06 37624.00 0.90833 1/3/03 10:06 pm 55.2

=B4&"/"&C4&"/"&"0"&D4 =E4&":"&F4 =DATEVALUE(G4) =TIMEVALUE(H4) =I4+J4

22:07 37624.00 0.921528 1/3/03 10:07 pm 55.3

22:09 37624.00 0.922917 1/3/03 10:09 pm 55.41/3/031 3 3 22 09

FIGURE 3. Manipulating a rough data set to provide a serial number that represents date and time for graphing purposes. The columnshighlighted in yellow were inserted into the original data set. Then, the formulas shown in the row highlighted in green were placed inthe cells in these new columns to generate the data that fills them. Column K was then formatted as a date and time, which caused thespreadsheet to display the serial number obtained by adding columns I and J as a date and time. Columns K and L can now be used to plota graph of temperature vs. date and time.

help you understand what is going on andpresent the results. However, when youdo this, bear in mind that the auto-scalingthat occurs when you use the chart wizardin this manner may hide some interestingdata. For example a duct pressure that ishunting 2 or 3 in. w.c. will probably looklike a flat line when it is plotted on thesame axis as with temperature and flowdata that may be auto scaled in terms of

hundreds, thousands or tens of thousandson the Y axis. Similar things can happenwith regard to the time frame of referencewhen a month’s worth of one minute datais compressed onto one graph.

FORMATTINGThere are several formatting tricks that

you can use to make your data more pre-sentable and do it quickly.

• If you click on a chart and copy it,and then click on another chart and se-lect “Paste Special” from the “Edit” dropdown menu and then pick the “Formats”button, you will paste most of the for-matting information from the first chartinto the second chart: things like fonts,colors, scaling etc.

• Sometimes, it is helpful to use similarcolors to graph related information. For

108 September 2003 • HPAC Engineering

E N E R G Y - E F F I C I E N T H O T E L S

Key Name: suffix Trend definition used

Point 1:

Point 2:

Point 3:

Point 4:

Point 5:

Point 6:

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Time interval: 5 min.

Date range: 08/29/2002 00:00:00 – 09/04/2002 23:59:59

Report timings: all hours

Time interval: 5 min.

Date range: 08/29/2002 00:00:00 – 09/04/2002 23:59:59

Report timings: all hours

Date Time Date and Time Point 1 Point 2 Point 3 Point 4 Point 5 Point 6

Point 1 Point 2 Point 3 Point 4 Point 5 Point 6

08/29/02

08/29/02

0:00:00

0:05:00

08/29/0212:00 am

08/29/0212:05 am

08/29/02

08/29/02

0:00:00

0:05:00

08/29/0212:00 am

08/29/0212:05 am

Off

Off

Off

Off

Off

Off

On

On

Off

Off

On

On

Before transposing

Key Trend definition used

Point 1:

Point 2:

Point 3:

Point 4:

Point 5:

Point 6:

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Trend COV (1.0000)

Date Date and TimeTime

B311.C01AP

Chiller 1 alarm

=TRANSPOSE(B2:B7)

=TRANSPOSE(C2:C7)

Off

Off

B311.C01EE

Chiller 1 enable

=TRANSPOSE(B2:B7)

=TRANSPOSE(C2:C7)

Off

Off

B311.C02AP

Chiller 2 alarm

=TRANSPOSE(B2:B7)

=TRANSPOSE(C2:C7)

Off

Off

B311.C02EE

Chiller 2 enable

=TRANSPOSE(B2:B7)

=TRANSPOSE(C2:C7)

On

On

B311.C03AP

Chiller 3 alarm

=TRANSPOSE(B2:B7)

=TRANSPOSE(C2:C7)

Off

Off

B311.C03EE

Chiller 3 enable

=TRANSPOSE(B2:B7)

=TRANSPOSE(C2:C7)

On

On

After transposing

Formulas

Formulas

B311.C01AP

B311.C01EE

B311.C02AP

B311.C02EE

B311.C03AP

B311.C03EE

Chiller 1 alarm

Chiller 1 enable

Chiller 2 alarm

Chiller 2 enable

Chiller 3 alarm

Chiller 3 enable

B311.C01AP

B311.C01EE

B311.C02AP

B311.C02EE

B311.C03AP

B311.C03EE

Chiller 1 alarm

Chiller 1 enable

Chiller 2 alarm

Chiller 2 enable

Chiller 3 alarm

Chiller 3 enable

1

2

3

4

5

6

7

8

9

10

11

12

13

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

A B C D E F G H I

Name: suffix

FIGURE 4. Before transposing, the columns in the data set only contained the generic labels Point_1, Point_2, etc. (topillustration). The transpose function allowed the numbers and labels in the vertical column to be displayed horizontally in tworows inserted below the generic labels. (For reference the formulas used to do this are show in the two rows in blue)

example, if you were graphing the mixedair temperature, preheat coil leaving airtemperature and chilled water coil leav-ing air temperature on the same graph asthe outputs that controlled them, youmight want to consider dark red, green,and blue for the temperatures and lightred, green, and blue for the related out-puts controlling them. For many people,the similar colors will help them correlatean input with its related output and as-similate the information more readily.

• Excel will allow you to add a secondaxis to your plots. This can be handywhen you are plotting things like processvalues along with the outputs controllingthem or when concurrently plotting datasets that have very different orders ofmagnitude (e.g., plotting duct static pres-sure along with flow rate). For instance, ifyou were plotting temperatures fromcoils along with the pneumatic pressuresthat were controlling the valves associatedwith the coils, you might consider plot-ting the valve data on a second axis with arange of 0 to 20 rather than on the sameaxis as the temperature data, which mayneed a range more along the lines of 0 to120 to show all of the data. Doing this al-lows you to better see the peaks and val-leys in the data, which are often useful di-agnostic indicators.

This approach can also be used to off-set the data strings so you can view themconcurrently but not on top of eachother. To continue the coil temperatureand valve output example, if all of thetemperatures were in the 40 to 110 Frange and the associated valve commandswere all between 3 and 15 psig, scalingthe temperature axis for 0 to 120 wouldput the temperature data higher on thegraph. You could then scale the pressuredata for 0 to 50 or 60. Since the pressuredata never exceeds 15 or 20, it will all ap-pear below the temperature data but willhave the peaks and valleys amplified morethan if they were simply plotted on thesame axis as the temperature parameters.Figure 5 demonstrates this concept ap-plied to display temperature differential

independentlyand magnifiedrelative to thetemperatures.

• One ofthe frustra-tions you mayexperiencewith Excel in-volves whathappens whenyou add cir-cles, lines, andnotes to agraph, thencopy and pasteit into a Worddocument.Sometimeswhen you dothis, the scal-ing becomescorrupt andthe lines, cir-cles and notesdo not end upin the correctplace when theimage is pasted into Word. There are acouple of ways that I have found to getaround this. One is to add the lines, ar-rows, and notes to the image after youpaste it into Word. Another is to createthe lines with equations rather than by in-serting a line. For example, if you want ahorizontal dashed line through the graphat 55 F, then insert a column in yourspreadsheet, copy and paste the value“55” into all of the cells, then select it as aseries to be included in the graph. You canthen format the series to give you the ef-fects you want such as color, dots, dashes,etc. If you don’t want it to show up in thelegend, then you can click on the legendand then click on the symbol for the lineand delete it.

CREATE USEFUL DATA FROM OTHER DATAAt first, this may sound like some of

the alleged practices of unscrupulous test-ing and balancing contractors. And,

while you could probably do somethinglike that, its not what I am advocating.Rather, I am suggesting that if you havesome good information in your data set,you can use it to perform calculationsthat will tell you more about the system.For example, if you know the enteringand leaving water temperature from achiller, and have some idea of the flow(which is often constant and easily deter-mined by a simple pump test or the bal-ance report data), then you can calculatetons for each sampling interval and usethis information to generate a load pro-file, and perhaps even ton-hours if youhave a lot of confidence in the accuracy ofthe data. If the relative calibration checkyou performed on your sensors beforeyou deployed them shows that they allread the same, then you probably canhave a high degree of confidence in thesecondary data you generate6 If there areapparent errors, then you may want to

109HPAC Engineering • September 2003

E N E R G Y - E F F I C I E N T H O T E L S

150140130120110100

908070605040302010

0

Outd

oor t

empe

ratu

re, F

and

pla

nt lo

ad,

perc

ent,

100

perc

ent =

2 c

hille

rs fu

lly lo

aded

Tem

pera

ture

diff

eren

ce, F

024681012141618

9/1312 am

9/133 am

9/136 am

9/139 am

9/1312 pm

Date and time

9/133 pm

9/136 pm

9/139 pm

9/1412 am

Chilled water plant load, percent(100 percent = two chillersfully loaded)Evaporator pump curcuittemperature difference, FOutdoor air trend line(moving averageto fill in for lost data)

B311.OSAT central plantoutside air temperatureDistribution circuittemperature difference, F

Figure 5 - Plant load profile generated from the distribution circuittemperature difference by assuming a constant flow through the chiller(determined from a pump test) and using the equation Load in btu/hrequals the flow in gpm times the difference between entering andleaving water temperature in degrees F. In this particular plot, thedifferential temperature data was also generated using math to allowus to understand the performance of the plant v.s. the load in that area.

apply corrections or limit your manipula-tions to more generalized values like per-cent full load instead of tons. The shape ofthe curve you generate can still providevaluable insights into the process even if itdoes not provide absolute accuracy.

Similar techniques can be used to makethe analysis of equipment sequencing eas-ier. For instance, if you are trying to un-derstand the sequencing of three parallelpumps, the raw data you get will typicallycontain only one of two values for eachpump, On or Off, or perhaps 1 or 0. Ifyou plot the three pumps on the samegraph, all of the lines will rest on top ofeach other making things hard to deci-pher. But, if you insert a couple ofcolumns and do some math, you canmake the values for On and Off be 1 and(-1) for the first pump, 2 and (-2) for the

second pump, 3 and (-3) for the thirdpump. Or, if you don’t care so muchabout which pump is running, just howmany, you can insert a column that sumsthe status values for the three pumps inquestion and generates a number be-tween 0 and 3 as an indication of howmany pumps are running.

CONCLUSIONThe bottom line is that the processing

power of modern spreadsheet softwarecan open many doors to you in terms ofenhancing your analysis capabilities onceyou have a delimited data set from yourdata logging equipment. So, if you findyourself faced with data that is not in theformat you had hoped for or you arewishing you had one more piece of infor-mation, spend a little time looking at the

possibilities available to you. (Clickingon the paste function button—the onethat looks like a lower case f with an xsubscript (fx ) on the standard Excel tool-bar usually gives you a pretty completelist of the possibilities.) You may find thatonce you get going, the real problem ismaking yourself stop.

END NOTES1) Sellers, D. (December, 2002). “Se-

lecting Data Loggers”HPAC Engineering.2) In most cases things will go faster if

you put the files you will be working withinto a folder on your hard disk.

3)It is a good idea to get in the habit ofalways saving before you do any majoroperation that could destroy your dataset in the spreadsheet program That way,you can abandon the results and reopenthe file without loosing anything. Thereare some operations that cannot be un-done but the "Undo" function.

4) Visit the Networked Controls mi-crosite at www.hpac.com to see a spread-sheet template that has the serial num-bers associated with 1 minute, 5 minutes,etc. and each hour of the day calculated.It also has a spot where you can enter adate and time and it converts it to a serialnumber. Its a useful tool when you arescaling an axis and saves having to calcu-late it each time.

Also available on the Networked Con-trols microsite at www.hpac.com is an ex-panded version of this article which in-cludes additional advice on what pointsshould be trended and how often to sam-ple them, as well as establishing a filenaming protocol to avoid confusion.

5) Sellers, D. (February 2003). “Data-logger Operation Tips” HPAC Engineer-ing.

6) Sellers, D.(March 2003). “Com-missiong Data Loggers,” HPAC Engi-neering.

For HPAC Engineering feature articlesdating back to January 1992, visitwww.hpac.com.

110 September 2003 • HPAC Engineering

E N E R G Y - E F F I C I E N T H O T E L S

Supplementing HVAC Data withWeather Data

Most control systems now have at least one outdoor air temperature sensor andincluding this information in the data set you pull for analyzing your HVAC process

can provide a lot of insight into what is going on. But, in many cases, outdoor tempera-ture only paints a part of the picture. Other parameters like relative humidity and dewpoint are necessary to fully understand what is really happening. If you have two psy-chrometric properties, you can use a psych chart to determine the others. Or for largedata sets, you can use the equations presented in the ASHRAE handbook of fundamen-tals to calculate some of the other parameters. However, if you don't have or only knowoutdoor air temperature or if you question the validity of the outdoor air moisture contentinformation you are getting from the system (or are math phobic) you may find yourselfwishing for information that is unavailable to you.

Fortunately, all hope is not lost. If you have an Internet connection and are willing tohunt around a little bit on the National Weather Service web sites, you can obtain hourlyweather data for any location with an automated weather station (virtually all airports ofany significant size). Typically, the report will include the date and time, air temperature,relative humidity, dew point, barometric pressure, wind direction and speed, precipita-tion, and an assessment of the general conditions (fair, cloudy, etc.)

If you want to know what was going on for the past 4 or 5 days, then you can pick it updirectly off of the web page; i.e. its for free. To retrieve the data, you just need to select it,copy it, and then paste it into a word processing document as text, save the document asa text file, and then import it into Excel as a delimited file. If you need older information,you can obtain it for a nominal fee from the regional climate center. The exact feestructure will vary from place to place, but for the Western Regional Climate Center that Iuse, it is; Service charge per request - $10, One month of hourly data (per location)- $12,One year of hourly data (per location) - $25. Turn around time ranges from a couple ofhours to a day. Data is available back through 1992, but a format change in 1996 makesthe data from that point in time forward much easier to use. You can get the data in a hardcopy format via the mail or a FAX or they can e-mail it to you as an ASCII file which willimport into Excel(r)