P R O C E S S E S

T HE EFFECT OF SURFACE ROUGHNESS

ON THE PERFORMANCE OF FINISHES.P ART 1. R O U G H N E S S C H A R A C T E R I Z A T I O N

AND STAIN PERFORMANCE

K LAUS R I C H T E R

W ILLIAM C. FE I S T

M ARK T. KNAEBE

A B S T R A C T

In this study, the relationship between the morphological structure of the outsidewood layer expressed as surface roughness, and the performance of coatings wasanalyzed. The surface roughness of five roughness categories (processed by planing,sanding, and bandsawing) on three wood substrates (vertical- and flat-grained westernredcedar and flat-grained southern yellow pine) was determined by stylus tracermeasurements. Several surface parameters were calculated to characterize the fiveroughness grades. Surface sanding proved to be an advantageous processing step priorto paint application. Sanded surfaces needed a relatively low quantity of paint forcoverage and showed best paint performances even on low-grade wood.

Wood structures exposed out-doors need protection against the influ-ence of sunlight and rain. Protection canbe achieved with a combination of build-ing design and efficient coating. Onebasic requirement for sufficient andlong-lasting paint performance is goodadhesion of the coating product on thewood surface. The ability of a woodsurface to accept and hold a paint coatingis determined by the natural charac-teristics of the wood species and themanufacturing processes used (3). Natu-ral factors (anatomical, physical, andchemical properties) vary considerably,not only between different species, buteven within the same species and tree.Their influence on paint performancecan only be predicted with a high rangeof variation and this influence is consid-ered to some extent in grading and se-lecting procedures.

But surface texture is not only deter-mined by the inherent morphologicalstructure of wood. According to a sur-

face-texture system proposed byMarian et al. (13), anatomic structurecauses a first-degree texture (e.g., tra-cheid or vessel diameter and cell wallthickness). A second-degree texture re-sults from the machining method itself(e.g., tooth marks from a saw and wavesformed by a machine planer). Third-de-gree texture results from variationwithin the machining method (e.g., vi-brations, misalignment, and dull tools).

There are two surface roughness tex-tures commonly used for wood sidingmaterials – smooth surface (planed) androughsawn. In exposure studies, the two

surface textures produce different per-formance results with finish systems.Penetrating stains and preservativetreatments gave better results on rough-sawn and flat-grained lumber (2,6,12)or rough-textured plywood (5). Thiswas a result of the substantially higherspreading rates generally achievable onrough substrates. Transparent finishesand white film-forming alkyd paintswere superior on smooth, edge-grainedsubstrates, because a more uniform filmthickness could be established, resultingin better moisture protection (3,12).Film-forming all-acrylic latex paintsshowed good performance and durabil-ity on both smooth and roughsawnwood (23).

However, the mechanisms responsi-ble for these characteristics are not yetfully understood. Surface texture wasnot characterized in any of these studies.In addition, no information could befound to indicate what roughness gradesare best for optimum durability and howfinish performance may vary on differ-ent roughness surfaces. This is contraryto the situation in wood adhesion sci-ence – an area with similar and compa-

The authors are, respectively, Research Scientist, EMPA, Überlandstrasse 129, CH-8600Dübendorf, Switzerland; retired Research Chemist; and Chemist, USDA Forest Serv., ForestProd. Lab., One Gifford Pinchot Dr., Madison, WI 53705-2398. The authors gratefullyacknowledge the support given by Tracey Duch and Peter G. Sotos for sample preparationand analysis. The use of trade or firm names in this publication is for reader information anddoes not imply endorsement by the EMPA or by the USDA of any product or service. Thispaper was received for publication in October 1994.

Forest Products Society 1995.Forest Prod. J. 45(7/8):91-97.

FOREST PRODUCTS JOURNAL VOL. 45, No. 7/8 91

rable problems of paint application andperformance, where much more atten-tion is given to surface preparation andits characterization prior to bonding(1,4,14,17).

The purpose of our investigation wasto determine how the roughness ofwood surfaces affects the overall per-formance of different coatings (16).This paper presents the results of sur-face characterization (roughness meas-urement and interpretation of roughnessand roughness standards) and the per-formance of stained samples in acceler-ated weathering. Subsequent paperswill discuss the relation of roughnessand paint adhesion and the performanceof painted and stained samples exposedoutdoors.

M A T E R I A L S A N D M E T H O D S

F I N I S H I N G S U B S T R A T E S

The wood species used were verti-cal- and flat-grained western redcedar(Thuja plicata) (WRC) and flat-grainedsouthern yellow pine (Pinus spp.)(SYP). The WRC was obtained directlyfrom a local lumberyard, where it wasavailable in a beveled form resultingfrom diagonal longitudinal bandsawcutting of the planed boards in the saw-mill, Thus, each board has a rough and asmooth (planed) surface. The SYP waspurchased several years ago and hadbeen stored since then in the laboratory.Its surface was originally planed.

The surface roughness categories(RC) listed in Table 1 were defined inrelation to the machining processesdone to the surface of the wood sam-ples. Sanding was done with a Solemdouble-belt sander, using only one beltand a 50-grit sandpaper. All sampleswere processed with the same feed rate;belt pressure was regulated automat-ically by a hydraulic device.

RCs 4 and 5 (Table 1) were proc-essed with a horizontal bandsaw

92

(Wood-Mizer LT 30). This type of sawallowed roughening the samples with aconstant feed rate and minimized un-controlled grooves in the surface pro-file, which were frequently found usingmanual feed. RC 4 was sawn with abandsaw with minimal tooth set (ap-proximately 0.04 mm) and a distancebetween two teeth of 19 mm (3/4 in.).The bandsaw used to produce RC 5 hada distance between two teeth of 22.3mm (7/8 in.) and a manually adjustedhigher set (approximately 0.06 mm).Both the bandsaw speed and feedingrate were maintained constant for thetwo RCs.

R O U G H N E S S M E A S U R E M E N T

Roughness measurements of all 366specimens prepared for finishing weredone with a commercial instrument(Perthometer S6P, drive unit PRK ofFeinprüf GmbH, 37008 Göttingen/Ger-many). This stylus tracing device wasdeveloped for quality control on workpieces with relatively smooth surfaces,such as metals and plastics. It was nec-essary to adjust the measurement rangeto scan the rough surfaces in our studyby elongating the length of the commer-cial pickup. Before being traced, allspecimens were conditioned to 12 per-cent equilibrium moisture content(EMC) in a climate room at 27°C and 65percent relative humidity (RH) for atleast 1 week. Table 2 lists the charac-teristics of the tracing process.

Because the number of data pointsmeasured per tracing unit (144points/mm) was more than necessaryand slowed later calculations, the datasets were compressed by selecting onlyevery sixth value. The remaining rawdata gave a detailed reproduction of thetotal movement of the stylus on thetraced surfaces, including the roughnessas well as the waviness and form of thesurface. The latter two componentswere excluded from the raw data profile

by a 2-step floating average of 100 datapoints each. The result was a de-trendedroughness profile representing 48 mm(reference length) of the tracing length.

Three standardized (DIN 4768/ISO4287) roughness parameters: averageroughness (Ra), average roughnessdepth (Rz), and maximum roughnessdepth (Rt); and two derived numbers:peak roughness (Pr) and peak index (Pi)were calculated and compared statisti-cally. Ra is the arithmetic mean of all,and Pr is the arithmetic mean of only thepeak and valley points within the refer-ence length. Rz measures maximumvertical distances within the referencelength. Pi was used in former studies toevaluate wood-based siding (11). Prwas modified by giving more emphasisto the larger peak and valley points.

F INISHING AND

PERFORMANCE RATING

A total of 66 boards (76 by 100 mm)were finished by brush with a linseedoil-based, semitransparent stain (cedarbrown), using two methods. First, thespecimens of all RCs were coated withthe recommended (normal) coverages.The finish applied per panel wasweighed and the spreading rates werecalculated. Then, duplicates of therougher samples (RCs 3,4, and 5) werefinished trying to apply only the quan-tity of stain used for the smooth surfaces(reduced coverage). This was achievedby using a 1:1 dilution of stain and min-eral spirits to duplicate the amount ofstain applied to the smooth surfaces.The resulting spreading rates are listedin Table 3.

The specimens were exposed to ac-celerated weathering in a xenon arcweathering chamber (Atlas Weather-O-Meter), where they received a daily cy-cle of 20 hours of light and 4 hours oflight plus water spray. The upper thirdsection of each specimen was protectedfrom weathering with a stainless steelplate, and served as a reference area.The finish performance (erosion anddiscoloration) was sated according to

JULY/AUGUST 1995

ASTM D 662-92 (erosion rate of semi-transparent stain) after 600, 1,200,1,800, and 2,400 hours of exposure. Thedegradation modes were rated on a 10 to1 scale where 10 = original conditionand 1 = total failure.

S T A T I S T I C A L E V A L U A T I O N

A two-way analysis of variance wascarried out on all data to determinewhether surface roughness (RC or se-lected roughness parameter) significantlyinfluenced the experimental data (spread-ing rate and paint performance). Differ-ences between the means of independentvariables were tested for significance us-ing Tukey’s Studentized Range Test. Sig-nificant differences were recorded at the 5percent probability level. All statistical

calculations were performed with theSAS software package (19).

R E S U L T S A N D D I S C U S S I O N

R O U G H N E S S E V A L U A T l O N

Stylus tracing was a suitable methodfor roughness determination. Althoughmainly used for polished and smoothsurfaces, the commercial device wasable to measure the coarse substratesafter some mechanical modifications.This is consistent with recent studieswhere the value of stylus tracing sys-tems for wood surface measurementswas shown (9, 18). The five calculatedroughness parameters selected werecompared statistically in a correlationanalysis, manifesting a high correlationbetween all parameters (Table 4). Thehighest and most homogeneous coeffi-

cients were found for Ra, representingthe arithmetic mean of the absolute val-ues of the profile deviation. Because itscalculation is standardized and the pa-rameter is used in other studies forroughness characterization (8, 15), Rawas selected in our study to quantifysurface roughness.

In Figure 1, Ra mean values and thestandard deviation of all 573 profilesscanned in our study (small boards weretraced once, larger ones twice) are plot-ted. The standard deviation was lowerfor the smooth samples where thesanded surfaces showed a better homo-geneity than the planed substrates.Sanding eliminates deviations better be-cause elements found in the anatomicstructure (such as resin ducts or early-wood/latewood differences) are morepronounced in the roughness profiles ofthe knife-planed surfaces, resulting inthe higher standard deviations.

The rougher surfaces were charac-terized by a much higher variabilitywithin the individual categories. TheTukey procedure proved that the differ-ences between the means, except forRCs 3 and 4, were significant from eachother. Thus, stylus tracing roughnessevaluation allowed the characterizationand splitting up between the surfaces offour roughness grades: RC 1, RC 2, RCs3 and 4, and RC 5. Factory processing(RC 3) and the wood Mizer rougheningwith the normal saw blades in the labo-ratory (RC 4) both gave similar surfacetopographies.

The effects of wood species andgrain orientation on surface topographyin the five RCs are shown in Figure 2,where the mean Ra values for the threesubsets are depicted (CF = WRC/flatgrained, CV = WRC/vertical grained,and PF = SYP/flat grained). To deter-mine the variability within each speciesgroup and roughness class, the standarderror of the mean was calculated. Con-netted bars within the roughness groupsmark the means that were not differentat the 95 percent significance level. Forthe planed surfaces (RC 1), all meanswere significantly different from eachother with the highest value for the flat-sawn WRC and the lowest roughnessfor the SYP. The flat-grained WRC wasa low-quality grade and the factoryplaning resulted in uneven and irregularsurfaces caused by the poor wood qual-ity and worn or dull knives. Addition-

FOREST PRODUCTS JOURNAL VOL. 45, NO. 7/8 93

Figure 1. — Mean values and standard deviations of average roughness (Ra)(expressed in micrometers) within the five roughness categories. The numbersquote the sample size. All means, except for roughness categories 3 and 4, weredifferent from each other.

against the cutting forces of the band-saw with normal set (RC 4) and is lesscompressed, so that the cutting qualityis more homogeneous. When the set ofthe blades is increased, as with the sec-ond bandsaw, more fibers are tom offthe earlywood tissue, whereas the late-wood zones are sheared off, resulting ina high roughness variability of the RC5-SYP surfaces.

R O U G H N E S S A N D

S P R E A D I N G R A T E

Ra and spreading rate numbers of the66 exposure panels subjected to acceler-ated weathering were compared in acorrelation analysis. The results mani-fest the inverse relationship betweensurface roughness and spreading ratesreported in previous studies (12,23).

An inverse relationship within eachof the three clusters presented in Figure3 exists between the roughness and thespreading rate with correlation coeffi-cients of r = -0.79 (smooth surfaces,normal application), r = -0.78 (roughsurfaces, normal application), and r =-0.42 (rough surfaces. reduced applica-

Figure 2. --- Mean values of average roughness (Ra) and the standard error ofmean for the species/grain subgroup within the five roughness categories. Barsconnected by lines beneath the graph within each roughness group marks meanvalues not different from each other at a significance level of 5 percent.

ally, the differences in earlywood/late- WRC specimen). Obviously, sanding aswood density caused higher swelling in well as factory rough sawing reducedthe tangentially cut latewood, so that the influences of grain and wood spe-some of the profiles of flatsawn wood ties. Within the WRC group, the lowerlooked like densitograms. The fact that standard deviation of the vertical grainthe SYP surfaces had the lowest rough- substrates shows that a homogeneousness was due to the fresh planing in the wood quality results in a more consis-laboratory right before roughness was tent surface quality even after sandingmeasured. and planing. Compared to WRC, rough

No significant differences between SYP showed significantly lower Rathe substrates were found within the RC means. It can be assumed that the higher2 and RC 3 and 4 categories (for the density SYP builds up more resistance

94

tion). The slope of the regression linesdepicts that within the samples paintedwith normal spread, the relationshipwas stronger for the smoother surfacescompared to the rougher samples.

As was intended by the study design,the reduced spreading rate on therougher surfaces generated significantlydifferent results from the normalspreading rates on similar rough sam-ples. The spreading rates ranged in thesame order as the coverages for thesmooth surfaces. The slope of the re-gression line is nearly the same as forthe surfaces painted with normal spread.Tukey groupings based on all datapoints showed there was no significantinfluence on spreading rate either bywood species (one exception) and grainorientation.

Figure 3 also allows the visualiza-tion of variation within the RCs. It washighest for the sanded specimen withspreading rates between 440 and 250ft.2/gal. In this group, the SYP speci-mens needed significantly more paintthan the WRC samples, 373 versus 268ft.2/gal., although no significant differ-ences in the roughness of the twosanded species group had been found(Fig. 2).

The sanding process seems to revealthe best possible individual charac-

JULY/AUGUST 1995

teristics of the substrates as they refer tothe ability to accept coatings. To a lesserextent, the same is true for the planedsurfaces, where the spreading rates forthe RC 1 surfaces were grouped be-tween 490 and 390 ft.2/gal. With in-creasing roughness, the deviation in thespreading rates became smaller. Obvi-ously, anatomic structural elements im-portant for the spreading rate (e.g., celldimensions and earlywood/latewooddistribution in growth rings) are over-lapped by roughness effects (e.g., in-creased surface area caused by valleys,crevices, and upstanding fiber bundles).

A C C E L E R A T E D W E A T H E R I N G

The results of stain erosion and dis-coloration evaluation after four 600-hour periods of accelerated weatheringare shown in Table 3. The numbers rep-resent the averages of three replicates.Afler 2,400 hours, only small differ-ences between erosion and discolora-tion were found. In Figure 4, the resultsfor vertical grain WRC erosion aregraphed as an example to visualize thestain performance during the test on thedifferent RCs.

All boards finished with a reducedspreading rate (SR 2) showed signifi-cantly lower ratings in both stain ero-sion and discoloration as compared tothe normally finished boards. The onlyexception was in the SYP specimenswith reduced coverage, where discol-oration had increased during the lastexposure period and was rated similar tothe normally finished boards. The fail-ure in erosion appeared after the firstexposure cycle and lasted until the endof the test, when most of the stain waseroded completely (Table 3).

Within the specimens with a mini-mum finish spread, no clear influence ofthe roughness group was visible. Thisbehavior clearly demonstrated that thegood performance of stain on rough-sawn surfaces seen in previous studies(5-7) is primarily due to the higherquantity of finish spread on the roughersurfaces and not an effect of the in-creased surface roughness. The rapiderosion indicates that rough wood, ifpainted insufficiently, might degradefaster because the moisture uptake andretention of roughened surfaces arehigher, as are swelling and shrinkagemovements in the loosened fibers.

On normally finished wood, thesanded surfaces (RC 2) showed similar

(SYP) or better performance thanroughsawn specimens. Sanded verticalgrain WRC had superior ratings (7.0 forboth erosion and discoloration), but thedifferences in the low-quality flat-grained WRC were surprisingly small(6.3 and 6.0, respectively).

SYP, a species with poorer paint-holding characteristics, was rated 5.0 inboth erosion and discoloration for RC 2,which is the same rating as for the RC 5surfaces. This result is particularly inter-esting because spreading rates of thesanded boards were 50 to 60 percenthigher than those of the bandsawn sam-ples. This indicates it is feasible to reachan equal (SYP) or improved long-termperformance (WRC) with less than half

of the quantity of stain used on roughwood, which was thought to be the bestsurface texture until now. Sandingwould make stain application more effi-cient, not only economically (less ma-terial use), but also from an ecologicalstandpoint (fewer volatile organicchemical emissions).

The reason for the improved stainperformance on sanded boards can beseen in the substrate conditions realizedby a slight sanding. A microscopicevaluation of all surface grades showedthat the stain uniformly covered thesanded specimens (Fig. 5). It has beenreported that on a microscopic level,sanding or abrasive planing causes sur-face and subsurface damage when com-

Figure 3. — Spreading rates of stain in relation to surface roughness.

Figure 4. — Erosion rating of vertically grained, stained, western redcedar afterfour 600-hour periods of accelerated weathering. The numbers identify the corre-spending spreading rates. a

p

FOREST PRODUCTS JOURNAL VOL. 45, NO. 7/8 95

red to knife planing (4,14,21). Surfacedamage (broken or crushed cell walls) isthe reason for a poor bonding quality inshear and adhesion tests. ,

According to our results, this phe-nomenon proved to be beneficial forfinish performance. We assume that theupstanding and lightly crushed early-wood fibers created an increased super-ficial surface and exerted adsorptionforces that prevented the stain frompenetrating in deeper wood zones. Fur-ther, normal anatomical penetrationpaths (e.g., wood ray tissue) had beencrushed so that an uptake of stain in

deeper cell layers was prevented. It isknown that sanding alters the cellularstructure so that no anatomical rough-ness (first-degree roughness) is detect-able (22). On the other hand, the denselatewood bands were roughened suffi-ciently so that an increased finish ad-sorption was possible in these zones,too.

Possibly, the lightly loosened andupstanding cell wall material provided amechanical reinforcement for the finishlayer when covered and immersed bythe stain. The finish was concentrated atthe sanded surface layer where it nearly

Figure 5.— Microphotograph of a flat-grained, sanded, western redcedar speci-mens (top) and a flat-grained, planed, southern yellow pine specimens (bottom)after 2,400 hours of accelerated weathering. The darker areas on the left werecovered in the weather-o-meter and allow a comparison to the exposed surfaces.

forms a uniform film and gives protec-tion against radiation and water Thissupports earlier results, which reporteda complete masking of grain in paintedabrasive-planed samples after acceler-ated weathering when compared toknife-planed samples (10). Accordingto Gaby’s study, knife planers or match-ers may cause physical surface changes,especially on flatsawn southern pine,which subsequently can affect paintperformance, whereas abrasive planersdo not compress or burn the surface inthe way that may adversely affect paintand stain life,

The reinforcement also appears onthe bandsawn surfaces, but the higherroughness allows more penetration inthe low-density earlywood cells, whichstabilized its weak thin cell walls, butresulted in higher finish uptakes. Theselow-density earlywood zones becamethe starting points for the severe failureon planed flatsawn surfaces because thesurfaces were not covered and protectedsufficiently with finish substance.

It will be interesting to evaluate theboards installed at the exposure site inMadison, Wis., to see if the positiveinteraction of sanded surface and finishcharacteristics is limited to the stainedsamples or if painted boards will show asimilar performance. Because of the dif-ferent viscosities and wetting charac-teristics of stain and paints, it might bereasonable that another roughness grademay perform better. Conversely, allacrylic paint systems showed the bestperformance on both roughsawn andscratch-sanded boards after 7 years inoutdoor exposure at the USDA ForestService’s test fence (5), so that anequally good behavior might be ex-pected from painted specimens. How-ever, the approach outlined in this studyshould be replicated in additional stud-ies using other grit sizes and sandingprocesses to test their influence on fin-ish performance.

C O N C L U S I O N S

Surface roughness produced bysanding or sawing can affect the per-formance of finishes in several ways. Ithas been quantified that finish spreadingrates and surface roughness are relatedand that rough surface substrates needmore finish coverage per area thansmooth surface substrates. Very roughstained wood performed well in long-term exposure mainly because of the

96 JULY/AUGU5T 1995

high spread necessary to cover the sur-face, whereas rough wood finished withthe same amount of stain as applied tosmooth surfaces (limited amounts)failed completely after only short-timeweathering. The best stain performancewas found on sanded surfaces, with per-formance ratings even better than thosefound for very rough wood, but withless than half the amount of finish ap-plied.

In contrast to what is concluded innumerous studies on adhesive perform-ance, sanding seems to be a perfect sur-face preparation for coatings because itlevels off inherent differences in woodsurface properties resulting in an equaland homogeneous finish spread. Thelight roughening allows sufficient finishpenetration in dense latewood zonesand avoids an over penetration in early-wood tissue, areas that are usually se-vere failure zones on planed wood. Astabilization effect and reinforcementfor the finish layer by upstanding fibersand cell wall material is conceivable.

Surface sanding has proved to be anadvantageous processing step prior topaint application. Sanded surfacesneeded a relatively low quantity of paintfor coverage and showed best paint per-formances even on low-grade wood,which can improve the furture competi-tiveness of wood siding.

L I T E R A T U R E C I T E D

FOREST PRODUCTS JOURNAL VOL. 45, No. 7/8 97